Fractal Analysis

A Wikipedia Collection compiled by Loren Cobb

PDF generated using the open source mwlib toolkit. See http://code.pediapress.com/ for more information. PDF generated at: Thu, 08 Mar 2012 18:37:29 UTC

Contents
Articles
Fractal dimension Fractal analysis Box counting Multifractal system Entropy (information theory) Rnyi entropy Hausdorff dimension 1 8 10 14 19 30 33

References
Article Sources and Contributors Image Sources, Licenses and Contributors 39 40

Article Licenses
License 41

Fractal dimension

Fractal dimension

- 11.5 x 200 = 2300 km - 28 x 100 = 2800 km - 70 x 50 = 3500 km A fractal dimension is a ratio providing a statistical index of complexity comparing how detail in a pattern (strictly speaking, a fractal pattern) changes with the scale at which it is measured. It has also been characterized as a measure of the space-filling capacity of a pattern that tells how a fractal scales differently than the space it is embedded in; a fractal dimension is greater than the dimension of the space containing it and does not have to be an integer.[1][2][3] The essential idea of "fractured" dimensions has a long history in mathematics, but the term itself was brought to the fore by Benot Mandelbrot based on his 1967 paper on self-similarity in which he discussed fractional dimensions.[4] In that paper, Mandelbrot cited previous work by Lewis Fry Richardson describing the counter-intuitive notion that a coastline's measured length changes with the length of the measuring stick used (see Fig. 1). In terms of that notion, the fractal dimension of a coastline quantifies how the number of scaled measuring sticks required to measure the coastline changes with the scale applied to the stick.[5] There are several formal mathematical definitions of fractal dimension that build on this basic concept of change in detail with change in scale. One non-trivial example is the fractal dimension of a Koch snowflake. It has a topological dimension of 1, but it is by no means a rectifiable curve: the length of the curve between any two points on the Koch Snowflake is infinite. No small piece of it is line-like, but rather is composed of an infinite number of segments joined at different angles. The fractal dimension of a curve can be explained intuitively thinking of a fractal line as an object too detailed to be one-dimensional, but too simple to be two-dimensional[6]:3-4. Therefore its dimension might best be described not by its usual topological dimension of 1 but by its fractal dimension, which in this case is a number between one and two.

Fractal dimension

Introduction
A fractal dimension is an index for characterizing fractal patterns or sets by quantifying their complexity as a ratio of the change in detail to the change in scale.[5]:1 Several types of fractal dimension can be measured theoretically and empirically (see Fig. 2).[8][3] Fractal dimensions are used to characterize a broad spectrum of objects ranging from the abstract[9][3] to practical phenomena, including turbulence[5]:97-104, river networks:247-246, urban growth[10][11], human physiology[12][13], medicine[8], and market trends[14]. The essential idea of fractional or fractal dimensions has a long history in mathematics that can be traced back to the 1600s[5]:19[15], but the terms fractal and fractal dimension were coined by mathematician Benot Mandelbrot in 1975.[16] [8] [5] [9][2][14]

Fractal dimensions were first applied as an index characterizing complex geometric forms for which the details seemed more important than the gross picture.[16] For sets describing ordinary geometric shapes, the theoretical fractal dimension equals the set's familiar Euclidean or topological dimension. Thus, it is 0 for sets describing points (0-dimensional sets); 1 for sets describing lines (1-dimensional sets having length only); 2 for sets describing surfaces (2-dimensional sets having length and width); and 3 for sets describing volumes (3-dimensional sets having length, width, and height). But this changes for fractal sets. If the theoretical fractal dimension of a set exceeds its topological dimension, the set is considered to have fractal geometry.[17] Unlike topological dimensions, the fractal index can take non-integer values, indicating that a set fills its space qualitatively and quantitatively differently than an ordinary geometrical set does.[3][9][2] For instance, a curve with fractal dimension very near to 1, say 1.10, behaves quite like an ordinary line, but a curve with fractal dimension 1.9 winds convolutedly through space very nearly like a surface. Similarly, a surface with fractal dimension of 2.1 fills space very much like an ordinary surface, but one with a fractal dimension of 2.9 folds and flows to fill space rather nearly like a volume.[17]:48[18] This general relationship can be seen in the two images of fractal curves in Fig.2 and Fig. 3 - the 32-segment contour in Fig. 2, convoluted and space filling, has a fractal dimension of 1.67, compared to the perceptibly less complex Koch curve in Fig. 3, which has a fractal dimension of 1.26.

Figure 2. A 32-segment quadric fractal scaled and viewed through boxes of different sizes. The pattern illustrates self similarity. The theoretical fractal dimension for this fractal is log32/log8 = 1.67; its empirical fractal dimension from box [7] counting analysis is 1% :86 using fractal analysis software.

Fractal dimension

The relationship of an increasing fractal dimension with space-filling might be taken to mean fractal dimensions measure density, but that is not so; the two are not strictly correlated.[7] Instead, a fractal dimension measures complexity, a concept related to certain key features of fractals: self-similarity and detail or irregularity.[19] These features are Figure 3. The Koch curve is a classic iterated evident in the two examples of fractal curves. Both are curves with fractal curve. It is a theoretical construct that is topological dimension of 1, so one might hope to be able to measure made by iteratively scaling a starting segment. As their length or slope, as with ordinary lines. But we cannot do either of shown, each new segment is scaled by 1/3 into 4 new pieces laid end to end with 2 middle pieces these things, because fractal curves have complexity in the form of leaning toward each other between the other two [5] self-similarity and detail that ordinary lines lack. The self-similarity pieces, so that if they were a triangle its base lies in the infinite scaling, and the detail in the defining elements of would be the length of the middle piece, so that each set. The length between any two points on these curves is the whole new segment fits across the traditionally measured length between the undefined because the curves are theoretical constructs that never stop endpoints of the previous segment. Whereas the repeating themselves[20]. Every smaller piece is composed of an animation only shows a few iterations, the infinite number of scaled segments that look exactly like the first theoretical curve is scaled in this way infinitely. iteration. These are not rectifiable curves, meaning they cannot be Beyond about 6 iterations on an image this small, the detail is lost. measured by being broken down into many segments approximating their respective lengths. They cannot be characterized by finding their lengths or slopes. However, their fractal dimensions can be determined, which shows that both fill space more than ordinary lines but less than surfaces, and allows them to be compared in this regard. Note that the two fractal curves described above show a type of self-similarity that is exact with a repeating unit of detail that is readily visualized. This sort of structure can be extended to other spaces (e.g., a fractal that extends the Koch curve into 3-d space has a theoretical D=2.5849). However, such neatly countable complexity is only one example of the self-similarity and detail that are present in fractals.[3][14] The example of the coast line of Britain, for instance, exhibits self-similarity of an approximate pattern with approximate scaling.[5]:26 Overall, fractals show several types and degrees of self-similarity and detail that may not be easily visualized. These include, as examples, strange attractors for which the detail has been described as in essence, smooth portions piling up[17]:49, the Julia set, which can be seen to be complex swirls upon swirls, and heart rates, which are patterns of rough spikes repeated and scaled in time.[21] Fractal complexity may not always be resolvable into easily grasped units of detail and scale without complex analytic methods but it is still quantifiable through fractal dimensions.[5]:197; 262

History
The terms fractal dimension and fractal were coined by Mandelbrot in 1975[16], about a decade after he published his paper on self-similarity in the coastline of Britain. Various historical authorities credit him with also synthesizing centuries of complicated theoretical mathematics and engineering work and applying them in a new way to study complex geometries that defied description in usual linear terms.[22][15][23] The earliest roots of what Mandelbrot synthesized as the fractal dimension have been traced clearly back to writings about undifferentiable, infinitely self-similar functions, which are important in the mathematical definition of fractals, around the time that calculus was discovered in the mid 1600s[5]:405. There was a lull in the published work on such functions for a time after that, then a renewal starting in the late 1800s with the publishing of mathematical functions and sets that are today called canonical fractals (such as the eponymous works of von Koch[20], Sierpinski, and Julia), but at the time of their formulation were often considered antithetical mathematical "monsters".[23][15] These works were accompanied by perhaps the most pivotal point in the development of the concept of a fractal dimension through the work of Hausdorff in the early 1900s who defined a "fractional" dimension that has come to be named after him and is frequently invoked in defining modern fractals.[22] [4][5]:44 [17]

Fractal dimension See Fractal history for more information

Role of scaling
The concept of a fractal dimension rests in nonconventional views of scaling and dimension.[24] As Fig. 4 illustrates, traditional notions of geometry dictate that shapes scale predictably according to intuitive and familiar ideas about the space they are contained within, such that, for instance, measuring a line using first one measuring stick then another 1/3 its size, will give for the second stick a total length 3 times as many sticks long as with the first. This holds in 2 dimensions, as well. If one measures the area of a square then measures again with a box 1/3 the size of the original, one will find 9 times as many squares as with the first measure. Such familiar scaling relationships can be defined mathematically by the general scaling rule in Equation 1, where the variable stands for the number of new sticks, for the scaling factor, and for the fractal dimension:

Figure 4. Traditional notions of geometry for defining scaling and dimension.

(1)

This scaling rule typifies conventional rules about geometry and dimension - for lines, it quantifies that, because =3 when =1/3 as in the example above, =1, and for squares, because =9 when =1/3, =2. The same rule applies to fractal geometry but less intuitively. To elaborate, a fractal line measured at first to be one length, when remeasured using a new stick scaled by 1/3 of the old may not be the expected 3 but instead 4 times as many scaled sticks long. In this case, =4 when =1/3, and the value of can be found by rearranging Equation 1:

Figure 5. The first four iterations of the Koch snowflake, which has an approximate Hausdorff dimension of 1.2619.

(2)

That is, for a fractal described by

=4 when

=1/3,
[3]

=1.2619, a non-integer dimension that suggests the

fractal has a dimension not equal to the space it resides in. The scaling used in this example is the same scaling of the Koch curve and snowflake. Of note, these images themselves are not true fractals because the scaling described by the value of cannot continue infinitely for the simple reason that the images only exist to the point of their smallest component, a pixel. The theoretical pattern that the digital images represent, however, has no discrete pixel-like pieces, but rather is composed of an infinite number of infinitely scaled segments joined at different angles

Fractal dimension and does indeed have a fractal dimension of 1.2619. [24][5]

D is not a unique descriptor

As is the case with dimensions determined for lines, squares, and cubes, fractal dimensions are general descriptors that do not uniquely define patterns.[25][24] The value of D for the Koch fractal discussed above, for instance, quantifies the pattern's inherent scaling, but does not uniquely describe nor provide enough information to reconstruct it. Many fractal structures or patterns could be constructed that have the same scaling relationship but are dramatically different from the Koch curve, as is illustrated in Figure 6. For examples of how fractal patterns can be constructed, see Fractal, Sierpinski triangle, Mandelbrot set, Diffusion limited aggregation.

Figure 6. Two L-systems branching fractals that are made by producing 4 new parts for every 1/3 scaling so have the same theoretical as the Koch curve and for which the empirical box counting has been demonstrated with 2% [7] accuracy .

Examples
The concept of fractal dimension described in this article is a basic view of a complicated construct. The examples discussed here were chosen for clarity, and the scaling unit and ratios were known ahead of time. In practise, however, fractal dimensions can be determined using techniques that approximate scaling and detail from limits estimated from regression lines over log vs log plots of size vs scale. Several formal mathematical definitions of different types of fractal dimension are listed below. Although for some classic fractals all these dimensions coincide, in general they are not equivalent: Box counting dimension: D is estimated as the exponent of a power law.

Information dimension: D considers how the average information needed to identify an occupied box scales with box size; is a probability.

Correlation dimension D is based on

as the number of points used to generate a representation of a fractal and

g, the number of pairs of points closer than to each other.

Generalized or Rnyi dimensions The box-counting, information, and correlation dimensions can be seen as special cases of a continuous spectrum of generalized dimensions of order , defined by:

Multifractal dimensions: a special case of Rnyi dimensions where scaling behaviour varies in different parts of the pattern. Uncertainty exponent Hausdorff dimension Packing dimension Local connected dimension[26]

Fractal dimension

Estimating from real-world data

The fractal dimension measures described in this article are for formally-defined fractals. However, many real-world phenomena also exhibit limited or statistical fractal properties and fractal dimensions have been estimated for sampled data from many such phenomena using computer based fractal analysis techniques. Practical dimension estimates are affected by various methodological issues, and are sensitive to numerical or experimental noise and limitations in the amount of data. Nonetheless, the field is rapidly growing and as evidenced by searching databases such as PubMed[27], the past decade has seen methods develop from being largely theoretical to the point where estimated fractal dimensions for statistically self-similar phenomena have many practical applications in multifarious fields including diagnostic imaging[28][29], physiology[30], neuroscience[31], medicine,[32][33][34], physics[35][36], image analysis [37] [38], acoustics[39], Riemann zeta zeros[40] and electrochemical processes.[41].

Notes
[1] [2] [3] [4] Falconer, Kenneth (2003). Fractal Geometry. New York: Wiley. p.308. ISBN9780470848623. Sagan, Hans (1994). Space-Filling Curves. Berlin: Springer-Verlag. p.156. ISBN0387942653. Vicsek, Tams (1992). Fractal growth phenomena. Singapore New Jersey: World Scientific. p.10. ISBN9789810206680. Mandelbrot, B. (1967). "How Long is the Coast of Britain? Statistical Self-Similarity and Fractional Dimension". Science 156 (3775): 636638. doi:10.1126/science.156.3775.636. PMID17837158.

[5] Benot B. Mandelbrot (1983). The fractal geometry of nature (http:/ / books. google. com/ books?id=0R2LkE3N7-oC). Macmillan. ISBN978-0716711865. . Retrieved 1 February 2012. [6] Harte, David (2001). Multifractals. London: Chapman & Hall. ISBN9781584881544. [7] Karperien (2004). "Defining Microglial Morphology: Form, Function, and Fractal Dimension. Charles Sturt University. p.95. [8] Losa, Gabriele A.; Nonnenmacher, Theo F., eds. (2005). Fractals in biology and medicine (http:/ / books. google. com/ books?id=t9l9GdAt95gC). Springer. ISBN978-3-7643-7172-2. . Retrieved 1 February 2012. [9] Falconer, Kenneth (2003). Fractal Geometry. New York: Wiley. p.308. ISBN9780470848623. [10] Chen, Y. (2011). Hernandez Montoya, Alejandro Raul. ed. "Modeling Fractal Structure of City-Size Distributions Using Correlation Functions". PLoS ONE 6 (9): e24791. doi:10.1371/journal.pone.0024791. PMC3176775. PMID21949753. [11] "Applications" (http:/ / library. thinkquest. org/ 26242/ full/ ap/ ap. html). . Retrieved 2007-10-21. [12] Popescu, D. P.; Flueraru, C.; Mao, Y.; Chang, S.; Sowa, M. G. (2010). "Signal attenuation and box-counting fractal analysis of optical coherence tomography images of arterial tissue". Biomedical Optics Express 1 (1): 268277. doi:10.1364/boe.1.000268. PMC3005165. PMID21258464. [13] King, R. D.; George, A. T.; Jeon, T.; Hynan, L. S.; Youn, T. S.; Kennedy, D. N.; Dickerson, B.; the Alzheimers Disease Neuroimaging Initiative (2009). "Characterization of Atrophic Changes in the Cerebral Cortex Using Fractal Dimensional Analysis". Brain Imaging and Behavior 3 (2): 154166. doi:10.1007/s11682-008-9057-9. PMC2927230. PMID20740072. [14] Peters, Edgar (1996). Chaos and order in the capital markets : a new view of cycles, prices, and market volatility. New York: Wiley. ISBN0471139386. [15] Edgar, Gerald (2004). Classics on Fractals. Boulder: Westview Press. ISBN9780813341538. [16] Albers; Alexanderson (2008). "Benot Mandelbrot: In his own words". Mathematical people : profiles and interviews. Wellesley, Mass: AK Peters. p.214. ISBN9781568813400. [17] Mandelbrot, Benot (2004). Fractals and Chaos. Berlin: Springer. ISBN9780387201580. "A fractal set is one for which the fractal (Hausdorff-Besicovitch) dimension strictly exceeds the topological dimension" [18] See a graphic representation of different fractal dimensions [19] See Fractal characteristics [20] Helge von Koch, "On a continuous curve without tangents constructible from elementary geometry" In Gerald Edgar, ed. (2004). Classics on Fractals. Boulder: Westview Press. p.25-46. ISBN9780813341538. [21] Tan, C. O.; Cohen, M. A.; Eckberg, D. L.; Taylor, J. A. (2009). "Fractal properties of human heart period variability: Physiological and methodological implications". The Journal of Physiology 587 (15): 3929. doi:10.1113/jphysiol.2009.169219. [22] Gordon, Nigel (2000). Introducing fractal geometry. Duxford: Icon. p.71. ISBN9781840461237. [23] Trochet, Holly (2009). "A History of Fractal Geometry" (http:/ / www. webcitation. org/ 65DCT2znx). MacTutor History of Mathematics. Archived from the original on 4 February 2012. . Retrieved 4 February 2012. [24] Iannaccone, Khokha (1996). Fractal Geometry in Biological Systems. ISBN978-0849376368. [25] Vicsek, Tams (2001). Fluctuations and scaling in biology. Oxford [Oxfordshire]: Oxford University Press. ISBN0-19-850790-9. [26] Jelinek (2008). "Automated detection of proliferative retinopathy in clinical practice". Clinical Ophthalmology: 109122. doi:10.2147/OPTH.S1579. [27] "PubMed" (http:/ / www. ncbi. nlm. nih. gov/ pubmed?term=fractal dimension). Search terms fractal analysis, box counting, fractal dimension, multifractal. . Retrieved January 31,2012.

Fractal dimension
[28] Landini, G.; Murray, P. I.; Misson, G. P. (1995). "Local connected fractal dimensions and lacunarity analyses of 60 degrees fluorescein angiograms". Investigative ophthalmology & visual science 36 (13): 27492755. PMID7499097. [29] Cheng, Q. (1997). Mathematical Geology 29 (7): 919932. doi:10.1023/A:1022355723781. [30] Popescu, D. P.; Flueraru, C.; Mao, Y.; Chang, S.; Sowa, M. G. (2010). "Signal attenuation and box-counting fractal analysis of optical coherence tomography images of arterial tissue". Biomedical Optics Express 1 (1): 268277. doi:10.1364/boe.1.000268. PMC3005165. PMID21258464. [31] King, R. D.; George, A. T.; Jeon, T.; Hynan, L. S.; Youn, T. S.; Kennedy, D. N.; Dickerson, B.; the Alzheimers Disease Neuroimaging Initiative (2009). "Characterization of Atrophic Changes in the Cerebral Cortex Using Fractal Dimensional Analysis". Brain Imaging and Behavior 3 (2): 154166. doi:10.1007/s11682-008-9057-9. PMC2927230. PMID20740072. [32] Liu, J. Z.; Zhang, L. D.; Yue, G. H. (2003). "Fractal Dimension in Human Cerebellum Measured by Magnetic Resonance Imaging". Biophysical Journal 85 (6): 40414046. doi:10.1016/S0006-3495(03)74817-6. PMC1303704. PMID14645092. [33] Smith, T. G.; Lange, G. D.; Marks, W. B. (1996). "Fractal methods and results in cellular morphology dimensions, lacunarity and multifractals". Journal of Neuroscience Methods 69 (2): 123136. doi:10.1016/S0165-0270(96)00080-5. PMID8946315. [34] Li, J.; Du, Q.; Sun, C. (2009). "An improved box-counting method for image fractal dimension estimation". Pattern Recognition 42 (11): 2460. doi:10.1016/j.patcog.2009.03.001. [35] Dubuc, B.; Quiniou, J.; Roques-Carmes, C.; Tricot, C.; Zucker, S. (1989). "Evaluating the fractal dimension of profiles". Physical Review A 39 (3): 15001512. doi:10.1103/PhysRevA.39.1500. PMID9901387. [36] Roberts, A.; Cronin, A. (1996). "Unbiased estimation of multi-fractal dimensions of finite data sets". Physica A: Statistical Mechanics and its Applications 233 (34): 867. doi:10.1016/S0378-4371(96)00165-3. [37] Pierre Soille and Jean-F. Rivest (1996). "On the Validity of Fractal Dimension Measurements in Image Analysis" (http:/ / mdigest. jrc. ec. europa. eu/ soille/ soille-rivest96. pdf). Journal of Visual Communication and Image Representation 7 (3): 217229. doi:10.1006/jvci.1996.0020. ISSN1047-3203. . [38] Tolle, C. R.; McJunkin, T. R.; Gorsich, D. J. (2003). "Suboptimal minimum cluster volume cover-based method for measuring fractal dimension". IEEE Transactions on Pattern Analysis and Machine Intelligence 25: 32. doi:10.1109/TPAMI.2003.1159944. [39] Maragos, P.; Potamianos, A. (1999). "Fractal dimensions of speech sounds: Computation and application to automatic speech recognition". The Journal of the Acoustical Society of America 105 (3): 19251932. doi:10.1121/1.426738. PMID10089613. [40] Shanker, O. (2006). "Random matrices, generalized zeta functions and self-similarity of zero distributions". Journal of Physics A: Mathematical and General 39 (45): 13983. doi:10.1088/0305-4470/39/45/008. [41] Eftekhari, A. (2004). "Fractal Dimension of Electrochemical Reactions". Journal of the Electrochemical Society 151 (9): E291E296. doi:10.1149/1.1773583.

References Further Reading

Mandelbrot, Benot B., The (Mis)Behavior of Markets, A Fractal View of Risk, Ruin and Reward (Basic Books, 2004)

External links
(http://www.trusoft-international.com) TruSoft's Benot - Fractal Analysis Software product calculates fractal dimensions and hurst exponents. (http://www.stevec.org/fracdim/) Fractal Dimension Estimator Java Applet (http://rsb.info.nih.gov/ij/plugins/fraclac/FLHelp/Fractals.htm) Fractal Analysis Software for Biologists; free from NIH ImageJ website

Fractal analysis

Fractal analysis
Fractal analysis is assessing fractal characterisics of data. It consists of several methods to assign a fractal dimension and other fractal characteristics to a dataset which may be a theoretical dataset or a pattern or signal extracted from phenomena including natural geometric objects, sound, market fluctuations[1], heart rates[2], digital images[3], molecular motion, networks, etc. Fractal analysis is now widely used in all areas of science[4]. An important limitation of fractal analysis is that arriving at an empirically determined fractal dimension does not necessarily prove that a pattern is fractal; rather, other essential characteristics have to be considered.[5]

Types of fractal analysis

Several types of fractal analysis are done, including box counting, lacunarity analysis, mass methods, and multifractal analysis[5][1]. A common feature of all types of fractal analysis is the need for benchmark patterns against which to assess outputs.[6] These can be acquired with various types of fractal generating software capable of generating benchmark patterns suitable for this purpose, which generally differ from software designed to render fractal art.

Applications
Applications of fractal analysis include:[7]
[2] Heart rate analysis urban growth [8] Computer and video game design, especially computer graphics for organic environments and as part of procedural generation Fractography and fracture mechanics Fractal antennas Small size antennas using fractal shapes Small angle scattering theory of fractally rough systems T-shirts and other fashion

Diagnostic imaging Cancer research

[9]

neuroscience

[10][11] [13]

[12]

diagnostic imaging pathology [15][16]

Classification of histopathology slides in [14] medicine Fractal landscape or Coastline [5][17] complexity Enzyme/enzymology (Michaelis-Menten kinetics) An Equation of State for the Coexistence of [20] Stability and Flexibility in Proteins Generation of new music Generation of various art forms

geology

[18]

geography

[19] [21][22]

Generation of patterns for camouflage, such as MARPAT Digital sundial

archaeology

Seismology

[23][24]

Technical analysis of price series (see Elliott wave principle) Fractal analysis in music [26][27]

Wave propagation in self-similar [25] (fractal) media Soil studies [29]

search and rescue

[28]

Signal and image compression

Fractal analysis

References
[1] Peters, Edgar (1996). Chaos and order in the capital markets : a new view of cycles, prices, and market volatility. New York: Wiley. ISBN0471139386. [2] Tan, C. O.; Cohen, M. A.; Eckberg, D. L.; Taylor, J. A. (2009). "Fractal properties of human heart period variability: Physiological and methodological implications". The Journal of Physiology 587 (15): 3929. doi:10.1113/jphysiol.2009.169219. [3] Fractal Analysis of Digital Images http:/ / rsbweb. nih. gov/ ij/ plugins/ fraclac/ FLHelp/ Fractals. htm [4] Fractals: Complex Geometry, Patterns, and Scaling in Nature and Society (http:/ / www. worldscinet. com/ fractals/ fractals. shtml). ISSN1793-6543. . [5] Benot B. Mandelbrot (1983). The fractal geometry of nature (http:/ / books. google. com/ books?id=0R2LkE3N7-oC). Macmillan. ISBN978-0716711865. . Retrieved 1 February 2012. [6] Digital Images in FracLac, ImageJ [7] "Applications" (http:/ / library. thinkquest. org/ 26242/ full/ ap/ ap. html). . Retrieved 2007-10-21. [8] Chen, Y. (2011). Hernandez Montoya, Alejandro Raul. ed. "Modeling Fractal Structure of City-Size Distributions Using Correlation Functions". PLoS ONE 6 (9): e24791. doi:10.1371/journal.pone.0024791. PMC3176775. PMID21949753. [9] Karperien, A.; Jelinek, H. F.; Leandro, J. J.; Soares, J. V.; Cesar Jr, R. M.; Luckie, A. (2008). "Automated detection of proliferative retinopathy in clinical practice". Clinical ophthalmology (Auckland, N.Z.) 2 (1): 109122. PMC2698675. PMID19668394. [10] Karperien, A. L.; Jelinek, H. F.; Buchan, A. M. (2008). "Box-Counting Analysis of Microglia Form in Schizophrenia, Alzheimer's Disease and Affective Disorder". Fractals 16 (2): 103. doi:10.1142/S0218348X08003880. [11] Liu, J. Z.; Zhang, L. D.; Yue, G. H. (2003). "Fractal Dimension in Human Cerebellum Measured by Magnetic Resonance Imaging". Biophysical Journal 85 (6): 40414046. doi:10.1016/S0006-3495(03)74817-6. PMC1303704. PMID14645092. [12] Kam, Y.; Karperien, A.; Weidow, B.; Estrada, L.; Anderson, A. R.; Quaranta, V. (2009). "Nest expansion assay: A cancer systems biology approach to in vitro invasion measurements". BMC Research Notes 2: 130. doi:10.1186/1756-0500-2-130. PMC2716356. PMID19594934. [13] Karperien, A.; Jelinek, H. F.; Leandro, J. J.; Soares, J. V.; Cesar Jr, R. M.; Luckie, A. (2008). "Automated detection of proliferative retinopathy in clinical practice". Clinical ophthalmology (Auckland, N.Z.) 2 (1): 109122. PMC2698675. PMID19668394. [14] Losa, Gabriele A.; Nonnenmacher, Theo F., eds. (2005). Fractals in biology and medicine (http:/ / books. google. com/ books?id=t9l9GdAt95gC). Springer. ISBN978-3-7643-7172-2. . Retrieved 1 February 2012. [15] Smith, R. F.; Mohr, D. N.; Torres, V. E.; Offord, K. P.; Melton Lj, 3. (1989). "Renal insufficiency in community patients with mild asymptomatic microhematuria". Mayo Clinic proceedings. Mayo Clinic 64 (4): 409414. PMID2716356. [16] Landini, G. (2011). "Fractals in microscopy". Journal of Microscopy 241 (1): 18. doi:10.1111/j.1365-2818.2010.03454.x. PMID21118245. [17] Mandelbrot, B. (1967). "How Long is the Coast of Britain? Statistical Self-Similarity and Fractional Dimension". Science 156 (3775): 636638. doi:10.1126/science.156.3775.636. PMID17837158. [18] Cheng, Q. (1997). Mathematical Geology 29 (7): 919932. doi:10.1023/A:1022355723781. [19] Chen, Y. (2011). Hernandez Montoya, Alejandro Raul. ed. "Modeling Fractal Structure of City-Size Distributions Using Correlation Functions". PLoS ONE 6 (9): e24791. doi:10.1371/journal.pone.0024791. PMC3176775. PMID21949753. [20] http:/ / prl. aps. org/ abstract/ PRL/ v100/ i20/ e208101 [21] Burkle-Elizondo, G.; Valdez-Cepeda, R. D. (2006). "Fractal analysis of Mesoamerican pyramids". Nonlinear dynamics, psychology, and life sciences 10 (1): 105122. PMID16393505. [22] Brown, C. T.; Witschey, W. R. T.; Liebovitch, L. S. (2005). "The Broken Past: Fractals in Archaeology". Journal of Archaeological Method and Theory 12: 37. doi:10.1007/s10816-005-2396-6. [23] Vannucchi, P.; Leoni, L. (2007). "Structural characterization of the Costa Rica dcollement: Evidence for seismically-induced fluid pulsing". Earth and Planetary Science Letters 262 (34): 413. doi:10.1016/j.epsl.2007.07.056. [24] Didier Sornette (2004). Critical phenomena in natural sciences: chaos, fractals, self organization, and disorder : concepts and tools. Springer. pp.128140. ISBN9783540407546. [25] http:/ / fr. arxiv. org/ abs/ 0904. 0780 [26] http:/ / www. brotherstechnology. com/ math/ fractal-music. html [27] Brothers, H. J. (2007). "Structural Scaling in Bach's Cello Suite No. 3". Fractals 15: 8995. doi:10.1142/S0218348X0700337X. [28] Panteha Saeedi, and Soren A. Sorensen. "An Algorithmic Approach to Generate After-disaster Test Fields for Search and Rescue Agents" (http:/ / www. iaeng. org/ publication/ WCE2009/ WCE2009_pp93-98. pdf). Proceedings of The World Congress on Engineering 2009: 93-98. ISBN978-988-17012-5-1. . [29] Hu, S.; Cheng, Q.; Wang, L.; Xie, S. (2012). "Multifractal characterization of urban residential land price in space and time". Applied Geography 34: 161. doi:10.1016/j.apgeog.2011.10.016.

Fractal analysis

10

Further reading
Fractals and Fractal Analysis (http://rsb.info.nih.gov/ij/plugins/fraclac/FLHelp/Fractals.htm) Fractal analysis (http://www.fch.vutbr.cz/lectures/imagesci/download_ejournal/01_O.Zmeskal.pdf) Benoit - Fractal Analysis Software (http://www.trusoft.netmegs.com/) Fractal Analysis Methods for Human Heartbeat and Gait Dynamics (http://www.physionet.org/tutorials/fmnc/ index.shtml)

Box counting
Box counting is a method of gathering data for analyzing complex patterns by breaking a dataset, object, image, etc. into smaller and smaller pieces, typically "box"-shaped, and analyzing the pieces at each smaller scale. The essence of the process has been compared to zooming in or out using optical or computer based methods to examine how observations of detail change with scale. In box counting, however, rather than changing the magnification or resolution of a lens, the investigator changes the size of the element used to inspect the object or pattern (see Figure 1). Computer based box counting algorithms have been applied to patterns in 1-, 2-, and 3-dimensional spaces.[1][2] The technique is usually implemented in Figure 1. A 32-segment quadric fractal viewed through "boxes" of different sizes. The software for use on patterns extracted pattern illustrates self similarity. from digital media, although the fundamental method can be used to investigate some patterns physically. The technique arose out of and is used in fractal analysis. It also has application in related fields such as lacunarity and multifractal analysis.[3][4]

The method
Theoretically, the intent of box counting is to quantify fractal scaling, but from a practical perspective this would require that the scaling be known ahead of time. This can be seen in Figure 1 where choosing boxes of the right relative sizes readily shows how the pattern repeats itself at smaller scales. In fractal analysis, however, the scaling factor is not always known ahead of time, so box counting algorithms attempt to find an optimized way of cutting a pattern up that will reveal the scaling factor. The fundamental method for doing this starts with a set of measuring elementsboxesconsisting of an arbitrary number, called here for convenience, of sizes or calibres, which we will call the set of s. Then these -sized boxes are applied to the pattern and counted. To do this, for each in , a measuring element that is typically a 2-dimensional square or 3-dimensional box with side length corresponding to is used to scan a pattern or data set (e.g., an image or object) according to a predetermined scanning plan to

Box counting cover the relevant part of the data set, recording, i.e.,counting, for each step in the scan relevant features captured within the measuring element. [3][4]

11

The data
The relevant features gathered during box counting depend on the subject being investigated and the type of analysis being done. Two well-studied subjects of box counting, for instance, are binary (meaning having only two colours, usually black and white)[2] and gray-scale[5] digital images (i.e., jpgs, tiffs, etc.). Box counting is generally done on patterns extracted from such still images in which case the raw information recorded is typically based on Figure 2. The sequence above shows basic steps in features of pixels such as a predetermined colour value or range of extracting a binary contour pattern from an original colours or intensities. When box counting is done to determine a colour digital image of a neuron. fractal dimension known as the box counting dimension, the information recorded is usually either yes or no as to whether or not the box contained any pixels of the predetermined colour or range (i.e., the number of boxes containing relevant pixels at each is counted). For other types of analysis, the data sought may be the number of pixels that fall within the measuring box,[4] the range or average values of colours or intensities, the spatial arrangement amongst pixels within each box, or properties such as average speed (e.g., from particle flow).[6][7][5][8]

Scan types
Every box counting algorithm has a scanning plan that describes how the data will be gathered, in essence, how the box will be moved over the space containing the pattern. A variety of scanning strategies has been used in box counting algorithms, where a few basic approaches have been modified in order to address issues such as sampling, analysis methods, etc. Fixed grid scans The traditional approach is to scan in a non-overlapping regular grid or lattice pattern.[3][4] To illustrate, Figure 2a shows the typical pattern used in software that calculates box counting dimensions from patterns extracted into binary digital images of contours such as the fractal contour illustrated in Figure 1 or the classic example of the coastline of Britain often used to explain the method of finding a box counting dimension. The strategy simulates repeatedly laying a square box as though it were part of a grid overlaid on the image, such that the box for each never overlaps where it has previously been (see Figure 4). This is done until the entire area of interest has been scanned using each and the relevant information has been recorded.[9] [10] When used to find a box counting dimension, the method is modified to find an optimal covering.

Figure 2a. Boxes laid over an image as a fixed grid. Figure 2b. Boxes slid over an image in an overlapping pattern.Figure 2c. Boxes laid over an image concentrically focused on each pixel of interest.

Box counting

12

Sliding box scans Another approach that has been used is a sliding box algorithm, in which each box is slid over the image overlapping the previous placement. Figure 2b illustrates the basic pattern of scanning using a sliding box. The fixed grid approach can be seen as a sliding box algorithm with the increments horizontally and vertically equal to . Sliding box algorithms are often used for analyzing textures in lacunarity analysis and have also been applied to multifractal analysis[8][2][11][12][13] Subsampling and local dimensions Box counting may also be used to determine local variation as opposed to global measures describing an entire pattern. Local variation can be assessed after the data have been gathered and analyzed (e.g., some software colour codes areas according to the fractal dimension for each subsample), but a third approach to box counting is to move the box according to some feature related to the pixels of interest. In local connected dimension box counting algorithms, for instance, the box for each is centred on each pixel of interest, as illustrated in Figure 2c.[7]

Figure 3. Retinal vasculature revealed through box counting analysis; colour coded local connected fractal dimension analysis done with FracLac freeware for biological image analysis.

Methodological considerations
The implementation of any box counting algorithm has to specify certain details such as how to determine the actual values in ,

Figure 4. It takes 12 green but 14 yellow boxes to completely cover the black pixels in these identical images. The difference is attributable to the position of the grid, illustrating the importance of grid placement in box counting.

including the minimum and maximum sizes to use and the method of incrementing between sizes. Many such details reflect practical matters such as the size of a digital image but also technical issues related to the specific analysis that will be performed on the data. Another issue that has received considerable attention is how to approximate the so-called "optimal covering" for determining box counting dimensions and assessing multifractal scaling. [14][5][15][16]

Edge effects
One known issue in this respect is deciding what constitutes the edge of the useful information in a digital image, as the limits employed in the box counting strategy can affect the data gathered.

Scaling box size

The algorithm has to specify the type of increment to use between box sizes (e.g., linear vs exponential), which can have a profound effect on the results of a scan.

Grid orientation
As Figure 4 illustrates, the overall positioning of the boxes also influences the results of a box count. One approach in this respect is to scan from multiple orientations and use averaged or optimized data.[17][18] To address various methodological considerations, some software is written so users can specify many such details, and some includes methods such as smoothing the data after the fact to be more amenable to the type of analysis

Box counting being done.[19]

13

References
[1] Liu, J. Z.; Zhang, L. D.; Yue, G. H. (2003). "Fractal Dimension in Human Cerebellum Measured by Magnetic Resonance Imaging". Biophysical Journal 85 (6): 40414046. doi:10.1016/S0006-3495(03)74817-6. PMC1303704. PMID14645092. [2] Smith, T. G.; Lange, G. D.; Marks, W. B. (1996). "Fractal methods and results in cellular morphology dimensions, lacunarity and multifractals". Journal of Neuroscience Methods 69 (2): 123136. doi:10.1016/S0165-0270(96)00080-5. PMID8946315. [3] Mandelbrot (1983). The Fractal Geometry of Nature. ISBN978-0716711865. [4] Iannaccone, Khokha (1996). Fractal Geometry in Biological Systems. pp.143. ISBN978-0849376368. [5] Li, J.; Du, Q.; Sun, C. (2009). "An improved box-counting method for image fractal dimension estimation". Pattern Recognition 42 (11): 2460. doi:10.1016/j.patcog.2009.03.001. [6] Karperien, A.; Jelinek, H. F.; Leandro, J. J.; Soares, J. V.; Cesar Jr, R. M.; Luckie, A. (2008). "Automated detection of proliferative retinopathy in clinical practice". Clinical ophthalmology (Auckland, N.Z.) 2 (1): 109122. PMC2698675. PMID19668394. [7] Landini, G.; Murray, P. I.; Misson, G. P. (1995). "Local connected fractal dimensions and lacunarity analyses of 60 degrees fluorescein angiograms". Investigative ophthalmology & visual science 36 (13): 27492755. PMID7499097. [8] Cheng, Q. (1997). Mathematical Geology 29 (7): 919932. doi:10.1023/A:1022355723781. [9] Popescu, D. P.; Flueraru, C.; Mao, Y.; Chang, S.; Sowa, M. G. (2010). "Signal attenuation and box-counting fractal analysis of optical coherence tomography images of arterial tissue". Biomedical Optics Express 1 (1): 268277. doi:10.1364/boe.1.000268. PMC3005165. PMID21258464. [10] King, R. D.; George, A. T.; Jeon, T.; Hynan, L. S.; Youn, T. S.; Kennedy, D. N.; Dickerson, B.; the Alzheimers Disease Neuroimaging Initiative (2009). "Characterization of Atrophic Changes in the Cerebral Cortex Using Fractal Dimensional Analysis". Brain Imaging and Behavior 3 (2): 154166. doi:10.1007/s11682-008-9057-9. PMC2927230. PMID20740072. [11] Plotnick, R. E.; Gardner, R. H.; Hargrove, W. W.; Prestegaard, K.; Perlmutter, M. (1996). "Lacunarity analysis: A general technique for the analysis of spatial patterns". Physical review. E, Statistical physics, plasmas, fluids, and related interdisciplinary topics 53 (5): 54615468. PMID9964879. [12] Plotnick, R. E.; Gardner, R. H.; O'Neill, R. V. (1993). "Lacunarity indices as measures of landscape texture". Landscape Ecology 8 (3): 201. doi:10.1007/BF00125351. [13] McIntyre, N. E.; Wiens, J. A. (2000). Landscape Ecology 15 (4): 313. doi:10.1023/A:1008148514268. [14] Gorski, A. Z.; Skrzat, J. (2006). "Error estimation of the fractal dimension measurements of cranial sutures". Journal of Anatomy 208 (3): 353359. doi:10.1111/j.1469-7580.2006.00529.x. PMC2100241. PMID16533317. [15] Chhabra, A.; Jensen, R. V. (1989). "Direct determination of the f( alpha ) singularity spectrum". Physical review letters 62 (12): 13271330. PMID10039645. [16] Fernndez, E.; Bolea, J. A.; Ortega, G.; Louis, E. (1999). "Are neurons multifractals?". Journal of neuroscience methods 89 (2): 151157. PMID10491946. [17] Karperien (2004). Defining Microglial Morphology: Form, Function, and Fractal Dimension. Charles Sturt University, Australia. [18] Schulze, M. M.; Hutchings, N.; Simpson, T. L. (2008). "The Use of Fractal Analysis and Photometry to Estimate the Accuracy of Bulbar Redness Grading Scales". Investigative Ophthalmology & Visual Science 49 (4): 1398. doi:10.1167/iovs.07-1306. [19] Karperien (2002), Box Counting, http:/ / rsb. info. nih. gov/ ij/ plugins/ fraclac/ FLHelp/ BoxCounting. htm#sampling

Multifractal system

14

Multifractal system
A multifractal system is a generalization of a fractal system in which a single exponent (the fractal dimension) is not enough to describe its dynamics; instead, a continuous spectrum of exponents (the so-called singularity spectrum) is needed.[1] Multifractal systems are common in nature, especially geophysics. They include fully developed turbulence, stock market time series, real world scenes, the Suns magnetic field time series, heartbeat dynamics, human gait, and natural luminosity time series. Models have been proposed in various contexts ranging from turbulence in fluid dynamics to internet traffic, finance, image modeling, texture synthesis, meteorology, geophysics and more. From a practical perspective, multifractal analysis uses the A Strange Attractor that exhibits multifractal scaling mathematical basis of multifractal theory to investigate datasets, often in conjunction with other methods of fractal analysis and lacunarity analysis. The technique entails distorting datasets extracted from patterns to generate multifractal spectra that illustrate how scaling varies over the dataset. The techniques of multifractal analysis have been applied in a variety of practical situations such as predicting earthquakes and interpreting medical images.[2][3][4]

Definition
In a multifractal system , the behavior around any point is described by a local power law:

The exponent

is called the singularity exponent, as it describes the local degree of singularity or regularity

around the point . The ensemble formed by all the points that share the same singularity exponent is called the singularity manifold of exponent h, and is a fractal set of fractal dimension D(h). The curve D(h) versus h is called the singularity spectrum and fully describes the (statistical) distribution of the variable . In practice, the multifractal behaviour of a physical system is not directly characterized by its singularity . Indeed, multifractal specrum D(h). Data analysis rather gives access to the multiscaling exponents

signals generally obey a scale invariance property which yields power law behaviours for multiresolution quantities depending on their scale . Depending on the object under study, these multiresolution quantities, denoted by in the following, can be local averages in boxes of size , gradients over distance , wavelet coefficients at scale ... For multifractal objects, one usually observes a global power law scaling of the form: . When such a behaviour is observed, one talks of

at least in some range of scales and for some range of orders scale invariance, self-similarity or multiscaling.
[5]

Multifractal system

15

Estimation
Using the so-called multifractal formalism, it can be shown that, under some well-suited assumptions, there exists a correspondence between the singularity spectrum and the multiscaling exponents through a Legendre transform. While the determination of calls for some exhaustive local analysis of the data, which would result relies on the use of statistical averages thanks are known, one can deduce an estimate of in difficult and numerically unstable calculations, the estimation of the and linear regressions in log-log diagrams. Once the

to a simple Legendre transform. Multifractal systems are often modeled by stochastic processes such as multiplicative cascades. Interestingly, the receives some statistical interpretation as they characterize the evolution of the distributions of the as goes from larger to smaller scales. This evolution is often called statistical intermittency and betrays a departure from Gaussian models. Modelling as a multiplicative cascade also leads to estimation of multifractal properties for relatively small datasets (Roberts & Cronin 1996). A maximum likelihood fit of a multiplicative cascade to the dataset not only estimates the complete spectrum, but also gives reasonable estimates of the errors (see the web service [6]).

Practical application of multifractal spectra

Multifractal analysis has been used in several fields in science to characterize various types of datasets[7]. In essence, multifractal analysis applies a distorting factor to datasets extracted from patterns, to compare how the data behave at each distortion. This is done using graphs known as multifractal spectra that illustrate how the distortions affect the data, analogous to viewing the dataset through a "distorting lens" as shown in the illustration[8]. Several types of multifractal spectra are used in practise.

Multifractal analysis is analogous to viewing a dataset through a series of distorting lenses to hone in on differences in scaling. The pattern shown is a Henon Map

Multifractal system

16

DQ vs Q
One practical multifractal spectrum is the graph of DQ vs Q, where DQ is the generalized dimension for a dataset and Q is an arbitrary set of exponents. The expression generalized dimension thus refers to a set of dimensions for a dataset (detailed calculations for determining the generalized dimension using box counting are described below). Dimensional ordering The general pattern of the graph of DQ vs Q can be used to assess the scaling in a pattern. The graph is generally decreasing, sigmoidal around Q=0, where D(Q=0) D(Q=1) D(Q=2). As illustrated in the figure, variation in this graphical spectrum can help distinguish patterns. The image shows D(Q) spectra from a multifractal analysis of binary images of non-, mono-, and multi-fractal sets. As is the case in the sample images, non- and mono-fractals tend to have flatter D(Q) spectra than multifractals.

DQ vs Q spectra for a non-fractal circle (empirical box counting dimension = 1.0), mono-fractal Quadric Cross (empirical box counting dimension = 1.49), and multifractal Henon Map (empirical box counting dimension = 1.29).

The generalized dimension also offers some important specific information. D(Q=0) is equal to the Capacity Dimension, which in the analysis shown in the figures here is the box counting dimension. D(Q=1) is equal to the Information Dimension, and D(Q=2) to the Correlation Dimension. This relates to the "multi" in multifractal whereby multifractals have multiple dimensions in the D(Q) vs Q spectra but monofractals stay rather flat in that area.[9][8]

vs
Another useful multifractal spectrum is the graph of vs (see calculations). These graphs generally rise to a maximum that approximates the fractal dimension at Q=0, and then fall. Like DQ vs Q spectra, they also show typical patterns useful for comparing non-, mono-, and multi-fractal patterns. In particular, for these spectra, nonand mono-fractals converge on certain values, whereas the spectra from multifractal patterns are typically humped over a broader extent.

Estimating multifractal scaling from box counting

Multifractal spectra can be determined from box counting on digital images. First, a box counting scan is done to determine how the pixels are distributed; then, this "mass distribution" becomes the basis for a series of calculations[9][10][8]. The chief idea is that for multifractals, the probability, , of a number of pixels, , appearing in a box, , varies as box size, , to some exponent, , which changes over the image, as in Eq.0.0. is calculated from the NB: For monofractals, in contrast, the exponent does not change meaningfully over the set. box counting pixel distribution as in Eq.2.0.
(Eq.0.0)

= an arbitrary scale (box size in box counting) at which the set is examined = the index for each box laid over the set for an = the number of pixels or mass in any box, , at size

= the total boxes that contained more than 0 pixels, for each
the total mass or sum of pixels in all boxes for this (Eq.1.0)

Multifractal system

17

the probability of this mass at

relative to the total mass for a box size

(Eq.2.0)

is used to observe how the pixel distribution behaves when distorted in certain ways as in Eq.3.0 and Eq.3.1: = an arbitrary range of values to use as exponents for distorting the data set
the sum of all mass probabilities distorted by being raised to this Q, for this box size (Eq.3.0)

When

, Eq.3.0 equals 1, the usual sum of all probabilities, and when .

, every term is equal to

1, so the sum is equal to the number of boxes counted,

how the distorted mass probability at a box compares to the distorted sum over all boxes at this box size

(Eq.3.1)

These distorting equations are further used to address how the set behaves when scaled or resolved or cut up into a series of -sized pieces and distorted by Q, to find different values for the dimension of the set, as in the following: An important feature of Eq.3.0 is that it can also be seen to vary according to scale raised to the exponent Eq.4.0:
(Eq.4.0)

in

Thus, a series of values for of for each

can be found from the slopes of the regression line for the log of Eq.3.0 vs the log

, based on Eq.4.1:
(Eq.4.1)

For the generalized dimension:

(Eq.5.0)

(Eq.5.1)

(Eq.5.2)

(Eq.5.3)

is estimated as the slope of the regression line for log A

,Q

vs log

where:

(Eq.6.0)

Then The mean

is found from Eq.5.3. is estimated as the slope of the log-log regression line for
(Eq.6.1)

vs

, where:

Multifractal system In practise, the probability distribution depends on how the dataset is sampled, so optimizing algorithms have been developed to ensure adequate sampling.[8]

18

References
[1] Harte, David (2001). Multifractals. London: Chapman & Hall. ISBN9781584881544. [2] Lopes, R.; Betrouni, N. (2009). "Fractal and multifractal analysis: A review". Medical Image Analysis 13 (4): 634649. doi:10.1016/j.media.2009.05.003. PMID19535282. [3] Moreno, P. A.; Vlez, P. E.; Martnez, E.; Garreta, L. E.; Daz, N. S.; Amador, S.; Tischer, I.; Gutirrez, J. M. et al (2011). "The human genome: A multifractal analysis". BMC Genomics 12: 506. doi:10.1186/1471-2164-12-506. PMID21999602. [4] Atupelage, C.; Nagahashi, H.; Yamaguchi, M.; Sakamoto, M.; Hashiguchi, A. (2012). "Multifractal feature descriptor for histopathology". Analytical cellular pathology (Amsterdam) 35 (2): 123126. doi:10.3233/ACP-2011-0045. PMID22101185. [5] A.J. Roberts and A. Cronin (1996). "Unbiased estimation of multi-fractal dimensions of finite data sets". Physica A 233: 867878. doi:10.1016/S0378-4371(96)00165-3. [6] http:/ / www. maths. adelaide. edu. au/ anthony. roberts/ multifractal. php [7] Trevino, J.; Liew, S. F.; Noh, H.; Cao, H.; Dal Negro, L. (2012). "Geometrical structure, multifractal spectra and localized optical modes of aperiodic Vogel spirals". Optics Express 20 (3): 3015. doi:10.1364/OE.20.003015. [8] Karperien, A (2002), What are Multifractals? (http:/ / rsbweb. nih. gov/ ij/ plugins/ fraclac/ FLHelp/ Multifractals. htm), ImageJ, , retrieved 2012-02-10 [9] Chhabra, A.; Jensen, R. (1989). "Direct determination of the f() singularity spectrum". Physical Review Letters 62 (12): 13271330. doi:10.1103/PhysRevLett.62.1327. PMID10039645. [10] Posadas, A. N. D.; Gimnez, D.; Bittelli, M.; Vaz, C. M. P.; Flury, M. (2001). "Multifractal Characterization of Soil Particle-Size Distributions". Soil Science Society of America Journal 65 (5): 1361. doi:10.2136/sssaj2001.6551361x.

External links
Stanley H.E., Meakin P. (1988). "Multifractal phenomena in physics and chemistry" (http://polymer.bu.edu/ hes/articles/sm88.pdf) (Review). Nature 335 (6189): 4059. doi:10.1038/335405a0. Alain Arneodo, et al. (2008). "Wavelet-based multifractal analysis" (http://www.scholarpedia.org/article/ Wavelet-based_multifractal_analysis). Scholarpedia 3 (3): 4103. doi:10.4249/scholarpedia.4103.

Entropy (information theory)

19

Entropy (information theory)

In information theory, entropy is a measure of the uncertainty associated with a random variable. In this context, the term usually refers to the Shannon entropy, which quantifies the expected value of the information contained in a message, usually in units such as bits. In this context, a 'message' means a specific realization of the random variable. Equivalently, the Shannon entropy is a measure of the average information content one is missing when one does not know the value of the random variable. The concept was introduced by Claude E. Shannon in his 1948 paper "A Mathematical Theory of Communication". Shannon's entropy represents an absolute limit on the best possible lossless compression of any communication, under certain constraints: treating messages to be encoded as a sequence of independent and identically-distributed random variables, Shannon's source coding theorem shows that, in the limit, the average length of the shortest possible representation to encode the messages in a given alphabet is their entropy divided by the logarithm of the number of symbols in the target alphabet. A single toss of a fair coin has an entropy of one bit. Two tosses has an entropy of two bits. The entropy rate for the coin is one bit per toss. However, if the coin is not fair, then the uncertainty is lower (if asked to bet on the next outcome, we would bet preferentially on the most frequent result), and thus the Shannon entropy is lower. Mathematically, a single coin flip (fair or not) is an example of a Bernoulli trial, and its entropy is given by the binary entropy function. A series of tosses of a two-headed coin will have zero entropy, since the outcomes are entirely predictable. The entropy rate of English text is between 1.0 and 1.5 bits per letter,[1] or as low as 0.6 to 1.3 bits per letter, according to estimates by Shannon based on human experiments.[2]

Introduction
Entropy is a measure of disorder, or more precisely unpredictability. For example, a series of coin tosses with a fair coin has maximum entropy, since there is no way to predict what will come next. A string of coin tosses with a coin with two heads and no tails has zero entropy, since the coin will always come up heads. Most collections of data in the real world lie somewhere in between. It is important to realize the difference between the entropy of a set of possible outcomes, and the entropy of a particular outcome. A single toss of a fair coin has an entropy of one bit, but a particular result (e.g. "heads") has zero entropy, since it is entirely "predictable". English text has fairly low entropy. In other words, it is fairly predictable. Even if we don't know exactly what is going to come next, we can be fairly certain that, for example, there will be many more e's than z's, or that the combination 'qu' will be much more common than any other combination with a 'q' in it and the combination 'th' will be more common than any of them. Uncompressed, English text has about one bit of entropy for each byte (eight bits) of message. If a compression scheme is losslessthat is, you can always recover the entire original message by uncompressingthen a compressed message has the same total entropy as the original, but in fewer bits. That is, it has more entropy per bit. This means a compressed message is more unpredictable, which is why messages are often compressed before being encrypted. Roughly speaking, Shannon's source coding theorem says that a lossless compression scheme cannot compress messages, on average, to have more than one bit of entropy per bit of message. The entropy of a message is in a certain sense a measure of how much information it really contains. Shannon's theorem also implies that no lossless compression scheme can compress all messages. If some messages come out smaller, at least one must come out larger. In the real world, this is not a problem, because we are generally only interested in compressing certain messages, for example English documents as opposed to random bytes, or digital photographs rather than noise, and don't care if our compressor makes random messages larger.

Entropy (information theory)

20

Definition
Named after Boltzmann's H-theorem, Shannon denoted the entropy H of a discrete random variable X with possible values {x1, ..., xn} as, Here E is the expected value, and I is the information content of X. I(X) is itself a random variable. If p denotes the probability mass function of X then the entropy can explicitly be written as

where b is the base of the logarithm used. Common values of b are 2, Euler's number e, and 10, and the unit of entropy is bit for b=2, nat for b=e, and dit (or digit) for b=10.[3] In the case of pi=0 for some i, the value of the corresponding summand 0logb0 is taken to be 0, which is consistent with the limit: . The proof of this limit can be quickly obtained applying l'Hpital's rule.

Example
Consider tossing a coin with known, not necessarily fair, probabilities of coming up heads or tails. The entropy of the unknown result of the next toss of the coin is maximized if the coin is fair (that is, if heads and tails both have equal probability 1/2). This is the situation of maximum uncertainty as it is most difficult to predict the outcome of the next toss; the result of each toss of the coin delivers a full 1 bit of information. However, if we know the coin is not fair, but comes up heads or tails with probabilities p and q, then there is less uncertainty. Every time it is tossed, one side is more likely to come up than the other. The reduced uncertainty is quantified in a lower entropy: on average each toss of the coin delivers less than a full 1 bit of information. The extreme case is that of a double-headed coin that never comes up tails, or a double-tailed coin that never results in a head. Then there is no uncertainty. The entropy is zero: each toss of the coin delivers no information. In this respect, entropy can by normalized by dividing it by information length. The measure is called metric entropy and allowed to measure the randomness of the information.

Entropy H(X) (i.e. the expected surprisal) of a coin flip, measured in bits, graphed versus the fairness of the coin Pr(X=1), where X=1 represents a result of heads. Note that the maximum of the graph depends on the distribution. Here, at most 1 bit is required to communicate the outcome of a fair coin flip (2 possible values), but the result of a fair die (6 possible values) would require at least log26 bits.

Rationale
For a random variable with outcomes , is defined as , the Shannon entropy, a measure of uncertainty (see further below) and denoted by

Entropy (information theory)

21
(1)

where

is the probability mass function of outcome

. possible outcomes (events) values, from to . The ,

To understand the meaning of Eq.(1), first consider a set of with equal probability uncertainty for such a set of outcomes is defined by

. An example would be a fair die with

(2)

The logarithm is used to provide the additivity characteristic for independent uncertainty. For example, consider appending to each value of the first die the value of a second die, which has possible outcomes . There are thus possible outcomes . The uncertainty for such a set of outcomes is then
(3)

Thus the uncertainty of playing with two dice is obtained by adding the uncertainty of the second die the uncertainty of the first die can write . Now return to the case of playing with one die only (the first one). Since the probability of each event is

to , we

In the case of a non-uniform probability mass function (or density in the case of continuous random variables), we let
(4)

which is also called a surprisal; the lower the probability surprise, i.e. , for the outcome , with . The average uncertainty

, i.e.

, the higher the uncertainty or the

being the average operator, is obtained by

(5)

and is used as the definition of the entropy

in Eq.(1). The above also explained why information entropy

and information uncertainty can be used interchangeably.[4] One may also define the conditional entropy of two events X and Y taking values xi and yj respectively, as

where p(xi,yj) is the probability that X=xi and Y=yj. This quantity should be understood as the amount of randomness in the random variable X given that you know the value of Y. For example, the entropy associated with a six-sided die is H(die), but if you were told that it had in fact landed on 1, 2, or 3, then its entropy would be equal to H(die: the die landed on 1, 2, or 3).

Entropy (information theory)

22

Aspects
Relationship to thermodynamic entropy
The inspiration for adopting the word entropy in information theory came from the close resemblance between Shannon's formula and very similar known formulae from thermodynamics. In statistical thermodynamics the most general formula for the thermodynamic entropy S of a thermodynamic system is the Gibbs entropy,

where kB is the Boltzmann constant, and pi is the probability of a microstate. The Gibbs entropy was defined by J. Willard Gibbs in 1878 after earlier work by Boltzmann (1872).[5] The Gibbs entropy translates over almost unchanged into the world of quantum physics to give the von Neumann entropy, introduced by John von Neumann in 1927,

where is the density matrix of the quantum mechanical system and Tr is the trace. At an everyday practical level the links between information entropy and thermodynamic entropy are not evident. Physicists and chemists are apt to be more interested in changes in entropy as a system spontaneously evolves away from its initial conditions, in accordance with the second law of thermodynamics, rather than an unchanging probability distribution. And, as the minuteness of Boltzmann's constant kB indicates, the changes in S / kB for even tiny amounts of substances in chemical and physical processes represent amounts of entropy which are so large as to be off the scale compared to anything seen in data compression or signal processing. Furthermore, in classical thermodynamics the entropy is defined in terms of macroscopic measurements and makes no reference to any probability distribution, which is central to the definition of information entropy. But, at a multidisciplinary level, connections can be made between thermodynamic and informational entropy, although it took many years in the development of the theories of statistical mechanics and information theory to make the relationship fully apparent. In fact, in the view of Jaynes (1957), thermodynamic entropy, as explained by statistical mechanics, should be seen as an application of Shannon's information theory: the thermodynamic entropy is interpreted as being proportional to the amount of further Shannon information needed to define the detailed microscopic state of the system, that remains uncommunicated by a description solely in terms of the macroscopic variables of classical thermodynamics, with the constant of proportionality being just the Boltzmann constant. For example, adding heat to a system increases its thermodynamic entropy because it increases the number of possible microscopic states for the system, thus making any complete state description longer. (See article: maximum entropy thermodynamics). Maxwell's demon can (hypothetically) reduce the thermodynamic entropy of a system by using information about the states of individual molecules; but, as Landauer (from 1961) and co-workers have shown, to function the demon himself must increase thermodynamic entropy in the process, by at least the amount of Shannon information he proposes to first acquire and store; and so the total thermodynamic entropy does not decrease (which resolves the paradox).

Entropy (information theory)

23

Entropy as information content

Entropy is defined in the context of a probabilistic model. Independent fair coin flips have an entropy of 1 bit per flip. A source that always generates a long string of B's has an entropy of 0, since the next character will always be a 'B'. The entropy rate of a data source means the average number of bits per symbol needed to encode it. Shannon's experiments with human predictors show an information rate of between 0.6 and 1.3 bits per character,[6] depending on the experimental setup; the PPM compression algorithm can achieve a compression ratio of 1.5 bits per character in English text. From the preceding example, note the following points: 1. The amount of entropy is not always an integer number of bits. 2. Many data bits may not convey information. For example, data structures often store information redundantly, or have identical sections regardless of the information in the data structure. Shannon's definition of entropy, when applied to an information source, can determine the minimum channel capacity required to reliably transmit the source as encoded binary digits (see caveat below in italics). The formula can be derived by calculating the mathematical expectation of the amount of information contained in a digit from the information source. See also Shannon-Hartley theorem. Shannon's entropy measures the information contained in a message as opposed to the portion of the message that is determined (or predictable). Examples of the latter include redundancy in language structure or statistical properties relating to the occurrence frequencies of letter or word pairs, triplets etc. See Markov chain.

Data compression
Entropy effectively bounds the performance of the strongest lossless (or nearly lossless) compression possible, which can be realized in theory by using the typical set or in practice using Huffman, Lempel-Ziv or arithmetic coding. The performance of existing data compression algorithms is often used as a rough estimate of the entropy of a block of data.[7][8] See also Kolmogorov complexity.

Limitations of entropy as information content

There are a number of entropy-related concepts that mathematically quantify information content in some way: the self-information of an individual message or symbol taken from a given probability distribution, the entropy of a given probability distribution of messages or symbols, and the entropy rate of a stochastic process. (The "rate of self-information" can also be defined for a particular sequence of messages or symbols generated by a given stochastic process: this will always be equal to the entropy rate in the case of a stationary process.) Other quantities of information are also used to compare or relate different sources of information. It is important not to confuse the above concepts. Oftentimes it is only clear from context which one is meant. For example, when someone says that the "entropy" of the English language is about 1.5 bits per character, they are actually modeling the English language as a stochastic process and talking about its entropy rate. Although entropy is often used as a characterization of the information content of a data source, this information content is not absolute: it depends crucially on the probabilistic model. A source that always generates the same symbol has an entropy rate of 0, but the definition of what a symbol is depends on the alphabet. Consider a source that produces the string ABABABABAB... in which A is always followed by B and vice versa. If the probabilistic model considers individual letters as independent, the entropy rate of the sequence is 1 bit per character. But if the sequence is considered as "AB AB AB AB AB..." with symbols as two-character blocks, then the entropy rate is 0 bits per character.

Entropy (information theory) However, if we use very large blocks, then the estimate of per-character entropy rate may become artificially low. This is because in reality, the probability distribution of the sequence is not knowable exactly; it is only an estimate. For example, suppose one considers the text of every book ever published as a sequence, with each symbol being the text of a complete book. If there are N published books, and each book is only published once, the estimate of the probability of each book is 1/N, and the entropy (in bits) is -log2 1/N = log2 N. As a practical code, this corresponds to assigning each book a unique identifier and using it in place of the text of the book whenever one wants to refer to the book. This is enormously useful for talking about books, but it is not so useful for characterizing the information content of an individual book, or of language in general: it is not possible to reconstruct the book from its identifier without knowing the probability distribution, that is, the complete text of all the books. The key idea is that the complexity of the probabilistic model must be considered. Kolmogorov complexity is a theoretical generalization of this idea that allows the consideration of the information content of a sequence independent of any particular probability model; it considers the shortest program for a universal computer that outputs the sequence. A code that achieves the entropy rate of a sequence for a given model, plus the codebook (i.e. the probabilistic model), is one such program, but it may not be the shortest. For example, the Fibonacci sequence is 1, 1, 2, 3, 5, 8, 13, ... . Treating the sequence as a message and each number as a symbol, there are almost as many symbols as there are characters in the message, giving an entropy of approximately log2(n). So the first 128 symbols of the Fibonacci sequence has an entropy of approximately 7 bits/symbol. However, the sequence can be expressed using a formula [F(n) = F(n-1) + F(n-2) for n={3,4,5,...}, F(1)=1, F(2)=1] and this formula has a much lower entropy and applies to any length of the Fibonacci sequence.

24

Limitations of entropy as a measure of unpredictability

In cryptanalysis, entropy is often roughly used as a measure of the unpredictability of a cryptographic key. For example, a 128-bit key that is randomly generated has 128 bits of entropy. It takes (on average) guesses to break by brute force. If the key's first digit is 0, and the others random, then the entropy is 127 bits, and it takes (on average) guesses. However, this measure fails if the possible keys are not of equal probability. If the key is half the time "password" and half the time a true random 128-bit key, then the entropy is approximately 65 bits. Yet half the time the key may be guessed on the first try, if your first guess is "password", and on average, it takes around guesses (not ) to break this password. Similarly, consider a 1000000-digit binary one-time pad. If the pad has 1000000 bits of entropy, it is perfect. If the pad has 999999 bits of entropy, evenly distributed (each individual bit of the pad having 0.999999 bits of entropy) it may still be considered very good. But if the pad has 999999 bits of entropy, where the first digit is fixed and the remaining 999999 digits are perfectly random, then the first digit of the ciphertext will not be encrypted at all.

Data as a Markov process

A common way to define entropy for text is based on the Markov model of text. For an order-0 source (each character is selected independent of the last characters), the binary entropy is:

where pi is the probability of i. For a first-order Markov source (one in which the probability of selecting a character is dependent only on the immediately preceding character), the entropy rate is:

where i is a state (certain preceding characters) and For a second order Markov source, the entropy rate is

is the probability of

given

as the previous character.

Entropy (information theory)

25

b-ary entropy
In general the b-ary entropy of a source = (S,P) with source alphabet S = {a1, ..., an} and discrete probability distribution P = {p1, ..., pn} where pi is the probability of ai (say pi = p(ai)) is defined by:

Note: the b in "b-ary entropy" is the number of different symbols of the "ideal alphabet" which is being used as the standard yardstick to measure source alphabets. In information theory, two symbols are necessary and sufficient for an alphabet to be able to encode information, therefore the default is to let b = 2 ("binary entropy"). Thus, the entropy of the source alphabet, with its given empiric probability distribution, is a number equal to the number (possibly fractional) of symbols of the "ideal alphabet", with an optimal probability distribution, necessary to encode for each symbol of the source alphabet. Also note that "optimal probability distribution" here means a uniform distribution: a source alphabet with n symbols has the highest possible entropy (for an alphabet with n symbols) when the probability distribution of the alphabet is uniform. This optimal entropy turns out to be .

Efficiency
A source alphabet with non-uniform distribution will have less entropy than if those symbols had uniform distribution (i.e. the "optimized alphabet"). This deficiency in entropy can be expressed as a ratio:

Efficiency has utility in quantifying the effective use of a communications channel.

Characterization
Shannon entropy is characterized by a small number of criteria, listed below. Any definition of entropy satisfying these assumptions has the form

where K is a constant corresponding to a choice of measurement units. In the following, and .

Continuity
The measure should be continuous, so that changing the values of the probabilities by a very small amount should only change the entropy by a small amount.

Symmetry
The measure should be unchanged if the outcomes xi are re-ordered. etc.

Maximum
The measure should be maximal if all the outcomes are equally likely (uncertainty is highest when all possible events are equiprobable).

Entropy (information theory)

26

For equiprobable events the entropy should increase with the number of outcomes.

Additivity
The amount of entropy should be independent of how the process is regarded as being divided into parts. This last functional relationship characterizes the entropy of a system with sub-systems. It demands that the entropy of a system can be calculated from the entropies of its sub-systems if the interactions between the sub-systems are known. Given an ensemble of n uniformly distributed elements that are divided into k boxes (sub-systems) with b1, b2, ... , bk elements each, the entropy of the whole ensemble should be equal to the sum of the entropy of the system of boxes and the individual entropies of the boxes, each weighted with the probability of being in that particular box. For positive integers bi where b1 + ... + bk = n,

Choosing k=n, b1 = ... = bn = 1 this implies that the entropy of a certain outcome is zero: This implies that the efficiency of a source alphabet with n symbols can be defined simply as being equal to its n-ary entropy. See also Redundancy (information theory).

Further properties
The Shannon entropy satisfies the following properties, for some of which it is useful to interpret entropy as the amount of information learned (or uncertainty eliminated) by revealing the value of a random variable X: Adding or removing an event with probability zero does not contribute to the entropy: . It can be confirmed using the Jensen inequality that . This maximal entropy of is effectively attained by a source alphabet having a uniform probability

distribution: uncertainty is maximal when all possible events are equiprobable. The entropy or the amount of information revealed by evaluating (X,Y) (that is, evaluating X and Y simultaneously) is equal to the information revealed by conducting two consecutive experiments: first evaluating the value of Y, then revealing the value of X given that you know the value of Y. This may be written as

If Y=f(X) where f is deterministic, then applying the previous formula to so ,

yields

thus the entropy of a variable can only decrease when the latter is passed through a deterministic function. If X and Y are two independent experiments, then knowing the value of Y doesn't influence our knowledge of the value of X (since the two don't influence each other by independence):

Entropy (information theory) The entropy of two simultaneous events is no more than the sum of the entropies of each individual event, and are equal if the two events are independent. More specifically, if X and Y are two random variables on the same probability space, and (X,Y) denotes their Cartesian product, then

27

Proving this mathematically follows easily from the previous two properties of entropy.

Extending discrete entropy to the continuous case: differential entropy

The Shannon entropy is restricted to random variables taking discrete values. The formula

where f denotes a probability density function on the real line, is analogous to the Shannon entropy and could thus be viewed as an extension of the Shannon entropy to the domain of real numbers. A precursor of the continuous entropy given in (1) is the expression for the functional in the H-theorem of

Boltzmann. Formula (1) is usually referred to as the continuous entropy, or differential entropy. Although the analogy between both functions is suggestive, the following question must be set: is the differential entropy a valid extension of the Shannon discrete entropy? Differential entropy lacks a number of properties that the Shannon discrete entropy has it can even be negative and thus corrections have been suggested, notably limiting density of discrete points. To answer this question, we must establish a connection between the two functions: We wish to obtain a generally finite measure as the bin size goes to zero. In the discrete case, the bin size is the (implicit) width of each of the n (finite or infinite) bins whose probabilities are denoted by pn. As we generalize to the continuous domain, we must make this width explicit. To do this, start with a continuous function f discretized as shown in the figure. As the figure indicates, by the mean-value theorem there exists a value xi in each bin such that

and thus the integral of the function f can be approximated (in the Riemannian sense) by

where this limit and "bin size goes to zero" are equivalent. We will denote

and expanding the logarithm, we have

As

, we have

Entropy (information theory) and also

28

But note that entropy:

as

, therefore we need a special definition of the differential or continuous

which is, as said before, referred to as the differential entropy. This means that the differential entropy is not a limit of the Shannon entropy for . Rather, it differs from the limit of the Shannon entropy by an infinite offset. It turns out as a result that, unlike the Shannon entropy, the differential entropy is not in general a good measure of uncertainty or information. For example, the differential entropy can be negative; also it is not invariant under continuous co-ordinate transformations. Another useful measure of entropy for the continuous case is the relative entropy of a distribution, defined as the Kullback-Leibler divergence from the distribution to a reference measure m(x),

The relative entropy carries over directly from discrete to continuous distributions, is always positive or zero, and is invariant under co-ordinate reparameterizations.

Use in combinatorics
Entropy has become a useful quantity in combinatorics.

Loomis-Whitney inequality
A simple example of this is an alternate proof of the Loomis-Whitney inequality: for every subset have , we

where the ith coordinate. The proof follows as a simple corollary of Shearer's inequality: if are subsets of

, that is,

is the orthogonal projection in

are random variables and

such that every integer between 1 and d lie in exactly r of these subsets, then

where

is the Cartesian product of random variables

with indexes j in

(so the dimension of this

vector is equal to the size of ). We sketch how Loomis-Whitney follows from this: Indeed, let X be a uniformly distributed random variable with values in A and so that each point in A occurs with equal probability. Then (by the further properties of entropy mentioned above) , where |A| denotes the cardinality of A. Let . The range of is contained in and hence

. Now use this to bound the right side of Shearer's inequality and exponentiate the opposite sides of the resulting inequality you obtain.

Entropy (information theory)

29

Approximation to binomial coefficient

For integers let . Then

where Here is a sketch proof. Note that

.[9] is one term of the expression

. Rearranging gives the upper bound. For the lower bound one first shows, using some algebra, that it is the largest term in the summation. But then, since there are approximately terms in the summation. Rearranging gives the lower bound. with exactly many 1's is .
[10]

A nice interpretation of this is that the number of binary strings of length

References
[1] Schneier, B: Applied Cryptography, Second edition, page 234. John Wiley and Sons. [2] Shannon, Claude E.: Prediction and entropy of printed English, The Bell System Technical Journal, 30:5064, January 1951. [3] Schneider, T.D, Information theory primer with an appendix on logarithms (http:/ / alum. mit. edu/ www/ toms/ paper/ primer/ primer. pdf), National Cancer Institute, 14 April 2007. [4] Jaynes, E.T. (May 1957). "Information Theory and Statistical Mechanics" (http:/ / bayes. wustl. edu/ etj/ articles/ theory. 1. pdf). Physical Review 106 (4): 620630. Bibcode1957PhRv..106..620J. doi:10.1103/PhysRev.106.620. . [5] Compare: Boltzmann, Ludwig (1896, 1898). Vorlesungen ber Gastheorie : 2 Volumes Leipzig 1895/98 UB: O 5262-6. English version: Lectures on gas theory. Translated by Stephen G. Brush (1964) Berkeley: University of California Press; (1995) New York: Dover ISBN 0-486-68455-5 [6] Mark Nelson (2006-08-24). "The Hutter Prize" (http:/ / marknelson. us/ 2006/ 08/ 24/ the-hutter-prize/ ). . Retrieved 2008-11-27. [7] T. Schrmann and P. Grassberger, Entropy Estimation of Symbol Sequences (http:/ / arxiv. org/ abs/ cond-mat/ 0203436), CHAOS,Vol. 6, No. 3 (1996) 414427 [8] T. Schrmann, Bias Analysis in Entropy Estimation (http:/ / arxiv. org/ abs/ cond-mat/ 0403192) J. Phys. A: Math. Gen. 37 (2004) L295-L301. [9] Aoki, New Approaches to Macroeconomic Modeling. page 43. [10] Probability and Computing, M. Mitzenmacher and E. Upfal, Cambridge University Press

This article incorporates material from Shannon's entropy on PlanetMath, which is licensed under the Creative Commons Attribution/Share-Alike License.

External links
Introduction to entropy and information (http://pespmc1.vub.ac.be/ENTRINFO.html) on Principia Cybernetica Web Entropy (http://www.mdpi.com/journal/entropy) an interdisciplinary journal on all aspect of the entropy concept. Open access. Information is not entropy, information is not uncertainty ! (http://alum.mit.edu/www/toms/information.is. not.uncertainty.html) a discussion of the use of the terms "information" and "entropy". I'm Confused: How Could Information Equal Entropy? (http://alum.mit.edu/www/toms/bionet.info-theory. faq.html#Information.Equal.Entropy) a similar discussion on the bionet.info-theory FAQ. Description of information entropy from "Tools for Thought" by Howard Rheingold (http://www.rheingold. com/texts/tft/6.html) A java applet representing Shannon's Experiment to Calculate the Entropy of English (http://math.ucsd.edu/ ~crypto/java/ENTROPY/) Slides on information gain and entropy (http://www.autonlab.org/tutorials/infogain.html)

Entropy (information theory) An Intuitive Guide to the Concept of Entropy Arising in Various Sectors of Science (http://en.wikibooks.org/ wiki/An_Intuitive_Guide_to_the_Concept_of_Entropy_Arising_in_Various_Sectors_of_Science) a wikibook on the interpretation of the concept of entropy. Calculator for Shannon entropy estimation and interpretation (http://www.shannonentropy.netmark.pl)

30

Rnyi entropy
In information theory, the Rnyi entropy, a generalisation of Shannon entropy, is one of a family of functionals for quantifying the diversity, uncertainty or randomness of a system. It is named after Alfrd Rnyi. The Rnyi entropy of order is defined for 0 and 1 as

where X is a discrete random variable, pi is the probability of the event {X =xi}, and the logarithm is base 2. If the probabilities are all the same then all the Rnyi entropies of the distribution are equal, with H(X) = log n. Otherwise the entropies are weakly decreasing as a function of . Higher values of , approaching infinity, give a Rnyi entropy which is increasingly determined by consideration of only the highest probability events. Lower values of , approaching zero, give a Rnyi entropy which increasingly weights all possible events more equally, regardless of their probabilities. The intermediate case =1 gives the Shannon entropy, which has special properties. When =0, it is the logarithm of the size of the support of X. The Rnyi entropies are important in ecology and statistics as indices of diversity. The Rnyi entropy is also important in quantum information, where it can be used as a measure of entanglement. In the Heisenberg XY spin chain model, the Rnyi entropy as a function of can be calculated explicitly by virtue of the fact that it is an automorphic function with respect to a particular subgroup of the modular group.[1][2]

H for some particular values of

Some particular cases:

which is the logarithm of the cardinality of X, sometimes called the Hartley entropy of X. In the limit that approaches 1, it can be shown using L'Hpital's Rule that converges to

which is the Shannon entropy. Collision entropy, sometimes just called "Rnyi entropy," refers to the case ,

where Y is a random variable independent of X but identically distributed to X. As

, the limit exists as

and this is called Min-entropy, because it is the smallest value of

Rnyi entropy

31

Inequalities between different values of

The two latter cases are related by . On the other hand the Shannon entropy can be arbitrarily high for a random variable X with fixed min-entropy. is because is because . .

since according to Jensen's inequality

Rnyi divergence
As well as the absolute Rnyi entropies, Rnyi also defined a spectrum of divergence measures generalising the KullbackLeibler divergence. The Rnyi divergence of order , where > 0, from a distribution P to a distribution Q is defined to be:

Like the Kullback-Leibler divergence, the Rnyi divergences are non-negative for >0. This divergence is also known as the alpha-divergence ( -divergence). Some special cases: : minus the log probability under Q that pi>0; : minus twice the logarithm of the Bhattacharyya coefficient; : the Kullback-Leibler divergence; : the log of the expected ratio of the probabilities; : the log of the maximum ratio of the probabilities.

Why = 1 is special
The value = 1, which gives the Shannon entropy and the KullbackLeibler divergence, is special because it is only when =1 that one can separate out variables A and X from a joint probability distribution, and write:

for the absolute entropies, and

for the relative entropies. The latter in particular means that if we seek a distribution p(x,a) which minimizes the divergence of some underlying prior measure m(x,a), and we acquire new information which only affects the distribution of a, then the distribution of p(x|a) remains m(x|a), unchanged. The other Rnyi divergences satisfy the criteria of being positive and continuous; being invariant under 1-to-1 co-ordinate transformations; and of combining additively when A and X are independent, so that if p(A,X) = p(A)p(X), then

Rnyi entropy

32

and

The stronger properties of the = 1 quantities, which allow the definition of conditional information and mutual information from communication theory, may be very important in other applications, or entirely unimportant, depending on those applications' requirements.

Exponential families
The Rnyi entropies and divergences for an exponential family admit simple expressions (Nielsen & Nock, 2011)

and

where

is a Jensen difference divergence.

Footnotes
[1] Its, A. R.; Korepin, V. E. (2010). "Generalized entropy of the Heisenberg spin chain" (http:/ / www. springerlink. com/ content/ vn2qt54344320m2g/ ). Theoretical and Mathematical Physics (Springer) 164 (3): 11361139. doi:10.1007/s11232-010-0091-6. . Retrieved 07-Mar-2012. [2] Franchini, F.; Its, A.R., Korepin, V.E. (2008). "Rnyi entropy as a measure of entanglement in quantum spin chain" (http:/ / arxiv. org/ pdf/ 0707. 2534). Journal of Physics A: Mathematical and Theoretical (IOPScience) 41 (025302). doi:10.1088/1751-8113/41/2/025302. . Retrieved 07-Mar-2012.

References
A. Rnyi (1961). "On measures of information and entropy" (http://digitalassets.lib.berkeley.edu/math/ucb/ text/math_s4_v1_article-27.pdf). Proceedings of the 4th Berkeley Symposium on Mathematics, Statistics and Probability 1960. pp.547561. A. O. Hero, O.Michael and J. Gorman (2002). Alpha-divergences for Classification, Indexing and Retrieval (http://www.eecs.umich.edu/~hero/Preprints/cspl-328.pdf). F. Nielsen and S. Boltz (2010). "The Burbea-Rao and Bhattacharyya centroids". arXiv:1004.5049. Nielsen, Frank; Nock, Richard (2011). "On Rnyi and Tsallis entropies and divergences for exponential families". arXiv:1105.3259. Rosso, O.A., "EEG analysis using wavelet-based information tools", Journal of Neuroscience Methods, 153 (2006) 163182. T. van Erven (2010). When Data Compression and Statistics Disagree (Ph.D. thesis). hdl:1887/15879. Chapter 6 Frank Nielsen and Richard Nock (2012). "A closed-form expression for the Sharma-Mittal entropy of exponential families" (http://iopscience.iop.org/1751-8121/45/3/032003/). Journal of Physics A: Mathematical and Theoretical.

Hausdorff dimension

33

Hausdorff dimension
In mathematics, the Hausdorff dimension (also known as the HausdorffBesicovitch dimension) is an extended non-negative real number associated with any metric space. The Hausdorff dimension generalizes the notion of the dimension of a real vector space. That is, the Hausdorff dimension of an n-dimensional inner product space equals n. This means, for example, the Hausdorff dimension of a point is zero, the Hausdorff Estimating the Hausdorff dimension of the coast of Great Britain dimension of a line is one, and the Hausdorff dimension of the plane is two. There are, however, many irregular sets that have noninteger Hausdorff dimension. The concept was introduced in 1918 by the mathematician Felix Hausdorff. Many of the technical developments used to compute the Hausdorff dimension for highly irregular sets were obtained by Abram Samoilovitch Besicovitch.

Intuition
The intuitive dimension of a geometric object is the number of independent parameters you need to pick out a unique point inside. But you can easily take a single real number, one parameter, and split its digits to make two real numbers. The example of a space-filling curve shows that you can even take one real number into two continuously, so that a one-dimensional object can completely fill up a higher dimensional object. Every space filling curve hits every point Sierpinski triangle. A space with fractal dimension log 3 / log 2, which is many times, and does not have a continuous approximately 1.5849625 inverse. It is impossible to map two dimensions onto one in a way that is continuous and continuously invertible. The topological dimension explains why. The Lebesgue covering dimension is defined as the minimum number of overlaps that small open balls need to have in order to completely cover the object. When you try to cover a line by dropping intervals on it, you always end up covering some points twice. Covering a plane with disks, you end up covering some points three times, etc. The topological dimension tells you how many different little balls connect a given point to other points in the space, generically. It tells you how difficult it is to break a geometric object apart into pieces by removing slices. But the topological dimension doesn't tell you anything about volumes. A curve which is almost space filling can still have topological dimension one, even if it fills up most of the area of a region. A fractal has an integer

Hausdorff dimension topological dimension, but in terms of the amount of space it takes up, it behaves as a higher dimensional space. The Hausdorff dimension defines the size notion of dimension, which requires a notion of radius, or metric. Consider the number N(r) of balls of radius at most r required to cover X completely. When r is small, N(r) is large. If N(r) always grows as 1/rd as r approaches zero, then X has Hausdorff dimension d. The precise definition requires that the dimension "d" so defined is a critical boundary between growth rates that are insufficient to cover the space, and growth rates that are overabundant. For shapes that are smooth, or shapes with a small number of corners, the shapes of traditional geometry and science, the Hausdorff dimension is an integer. But Benot Mandelbrot observed that fractals, sets with noninteger Hausdorff dimensions, are found everywhere in nature. He observed that the proper idealization of most rough shapes you see around you is not in terms of smooth idealized shapes, but in terms of fractal idealized shapes: clouds are not spheres, mountains are not cones, coastlines are not circles, and bark is not smooth, nor does lightning travel in a straight line. [1] The Hausdorff dimension is a successor to the less sophisticated but in practice very similar box-counting dimension or MinkowskiBouligand dimension. This counts the squares of graph paper in which a point of X can be found as the size of the squares is made smaller and smaller. For fractals that occur in nature, the two notions coincide. The packing dimension is yet another similar notion. These notions (packing dimension, Hausdorff dimension, MinkowskiBouligand dimension) all give the same value for many shapes, but there are well documented exceptions.

34

Formal definition
Let by be a metric space. If and , the -dimensional Hausdorff content of is defined

In other words, balls

is the infimum of the set of numbers covering with for each

such that there is some (indexed) collection of which satisfies is defined by such that the -dimensional such that the is empty the . (Here, we use

the standard convention that inf =.) The Hausdorff dimension of Equivalently, Hausdorff measure of may be defined as the infimum of the set of

is zero. This is the same as the supremum of the set of is infinite (except that when this latter set of numbers

-dimensional Hausdorff measure of Hausdorff dimension is zero).

Examples
The Euclidean space has Hausdorff dimension n. 1 The circle S has Hausdorff dimension 1. Countable sets have Hausdorff dimension 0. Fractals often are spaces whose Hausdorff dimension strictly exceeds the topological dimension. For example, the Cantor set (a zero-dimensional topological space) is a union of two copies of itself, each copy shrunk by a factor 1/3; this fact can be used to prove that its Hausdorff dimension is which is approximately The Sierpinski triangle is a union of three copies of itself, each copy shrunk by a factor of 1/2; this yields a Hausdorff dimension of , which is approximately . Space-filling curves like the Peano and the Sierpiski curve have the same Hausdorff dimension as the space they fill.

Hausdorff dimension The trajectory of Brownian motion in dimension 2 and above has Hausdorff dimension 2 almost surely. An early paper by Benoit Mandelbrot entitled How Long Is the Coast of Britain? Statistical Self-Similarity and Fractional Dimension and subsequent work by other authors have claimed that the Hausdorff dimension of many coastlines can be estimated. Their results have varied from 1.02 for the coastline of South Africa to 1.25 for the west coast of Great Britain. However, 'fractal dimensions' of coastlines and many other natural phenomena are largely heuristic and cannot be regarded rigorously as a Hausdorff dimension. It is based on scaling properties of coastlines at a large range of scales; however, it does not include all arbitrarily small scales, where measurements would depend on atomic and sub-atomic structures, and are not well defined. The bond system of an amorphous solid changes its Hausdorff dimension from Euclidian 3 below glass transition temperature Tg (where the amorphous material is solid), to fractal 2.550.05 above Tg, where the amorphous material is liquid.[2]

35

Properties of Hausdorff dimension

Hausdorff dimension and inductive dimension
Let X be an arbitrary separable metric space. There is a topological notion of inductive dimension for X which is defined recursively. It is always an integer (or +) and is denoted dimind(X). Theorem. Suppose X is non-empty. Then

Moreover

where Y ranges over metric spaces homeomorphic to X. In other words, X and Y have the same underlying set of points and the metric dY of Y is topologically equivalent to dX. These results were originally established by Edward Szpilrajn (19071976). The treatment in Chapter VII of the Hurewicz and Wallman reference is particularly recommended.

Hausdorff dimension and Minkowski dimension

The Minkowski dimension is similar to the Hausdorff dimension, except that it is not associated with a measure. The Minkowski dimension of a set is at least as large as the Hausdorff dimension. In many situations, they are equal. However, the set of rational points in has Hausdorff dimension zero and Minkowski dimension one. There are also compact sets for which the Minkowski dimension is strictly larger than the Hausdorff dimension.

Hausdorff dimensions and Frostman measures

If there is a measure defined on Borel subsets of a metric space and for every ball in such that , then and . holds for some constant Hausdorff dimension.

A partial converse is provided by Frostman's lemma. That article also discusses another useful characterization of the

Hausdorff dimension

36

Behaviour under unions and products

If is a finite or countable union, then

This can be verified directly from the definition. If and are metric spaces, then the Hausdorff dimension of their product satisfies[3]

This inequality can be strict. It is possible to find two sets of dimension 0 whose product has dimension 1.[4] In the opposite direction, it is known that when and are Borel subsets of , the Hausdorff dimension of is bounded from above by the Hausdorff dimension of discussed in Mattila (1995). plus the upper packing dimension of . These facts are

Self-similar sets
Many sets defined by a self-similarity condition have dimensions which can be determined explicitly. Roughly, a set E is self-similar if it is the fixed point of a set-valued transformation , that is (E) = E, although the exact definition is given below. Theorem. Suppose are contractive mappings on Rn with contraction constant rj < 1. Then there is a unique non-empty compact set A such that

The theorem follows from Stefan Banach's contractive mapping fixed point theorem applied to the complete metric space of non-empty compact subsets of Rn with the Hausdorff distance.[5] To determine the dimension of the self-similar set A (in certain cases), we need a technical condition called the open set condition on the sequence of contractions i which is stated as follows: There is a relatively compact open set V such that

where the sets in union on the left are pairwise disjoint. Theorem. Suppose the open set condition holds and each i is a similitude, that is a composition of an isometry and a dilation around some point. Then the unique fixed point of is a set whose Hausdorff dimension is s where s is the unique solution of [6]

Note that the contraction coefficient of a similitude is the magnitude of the dilation. We can use this theorem to compute the Hausdorff dimension of the Sierpinski triangle (or sometimes called Sierpinski gasket). Consider three non-collinear points a1, a2, a3 in the plane R and let i be the dilation of ratio 1/2 around ai. The unique non-empty fixed point of the corresponding mapping is a Sierpinski gasket and the dimension s is the unique solution of

Taking natural logarithms of both sides of the above equation, we can solve for s, that is:

Hausdorff dimension

37

The Sierpinski gasket is self-similar. In general a set E which is a fixed point of a mapping

is self-similar if and only if the intersections

where s is the Hausdorff dimension of E and

denotes Hausdorff measure. This is clear in the case of the

Sierpinski gasket (the intersections are just points), but is also true more generally: Theorem. Under the same conditions as the previous theorem, the unique fixed point of is self-similar.

The Hausdorff dimension theorem

The following theorem deals with existence of fractals with given Hausdorff dimension in Euclidean spaces [7]: Theorem. For any real -dimensional Euclidean space. and integer , there is a continuum fractals with Hausdorff dimension in

Historical references
A. S. Besicovitch (1929). "On Linear Sets of Points of Fractional Dimensions". Mathematische Annalen 101 (1): 161193. doi:10.1007/BF01454831. A. S. Besicovitch; H. D. Ursell (1937). "Sets of Fractional Dimensions". Journal of the London Mathematical Society 12 (1): 1825. doi:10.1112/jlms/s1-12.45.18. Several selections from this volume are reprinted in Edgar, Gerald A. (1993). Classics on fractals. Boston: Addison-Wesley. ISBN0-201-58701-7. See chapters 9,10,11 F. Hausdorff (March 1919). "Dimension und ueres Ma". Mathematische Annalen 79 (12): 157179. doi:10.1007/BF01457179.

Notes
[1] Mandelbrot, Benot (1982). The Fractal Geometry of Nature. Lecture notes in mathematics 1358. W. H. Freeman. ISBN0716711869. [2] M.I. Ojovan, W.E. Lee. (2006). "Topologically disordered systems at the glass transition" (http:/ / eprints. whiterose. ac. uk/ 1958/ ). J. Phys.: Condensed Matter 18 (50): 1150720. Bibcode2006JPCM...1811507O. doi:10.1088/0953-8984/18/50/007. . [3] Marstrand, J. M. (1954). "The dimension of Cartesian product sets". Proc. Cambridge Philos. Soc. 50 (3): 198202. doi:10.1017/S0305004100029236. [4] Falconer, Kenneth J. (2003). Fractal geometry. Mathematical foundations and applications. John Wiley & Sons, Inc., Hoboken, New Jersey. [5] Falconer, K. J. (1985). "Theorem 8.3". The Geometry of Fractal Sets. Cambridge, UK: Cambridge University Press. ISBN0-521-25694-1. [6] Tsang, K. Y. (1986). "Dimensionality of Strange Attractors Determined Analytically" (http:/ / prl. aps. org/ abstract/ PRL/ v57/ i12/ p1390_1). Phys. Rev. Lett. 57 (12): 13901393. doi:10.1103/PhysRevLett.57.1390. PMID10033437. . [7] Soltanifar, M.(2006) On A Sequence of Cantor Fractals, Rose Hulman Undergraduate Mathematics Journal, Vol 7, No 1, paper 9.

Hausdorff dimension

38

References
Dodson, M. Maurice; Kristensen, Simon (June 12, 2003). "Hausdorff Dimension and Diophantine Approximation". Fractal geometry and applications: a jubilee of Beno\^it Mandelbrot. Part, --347, Proc. Sympos. Pure Math., 72, Part , Amer. Math. Soc., Providence, RI, . 1 (305). arXiv:math/0305399. Hurewicz, Witold; Wallman, Henry (1948). Dimension Theory. Princeton University Press. E. Szpilrajn (1937). "La dimension et la mesure". Fundamenta Mathematica 28: 819. Marstrand, J. M. (1954). "The dimension of cartesian product sets". Proc. Cambridge Philos. Soc. 50 (3): 198202. doi:10.1017/S0305004100029236. Mattila, Pertti (1995). Geometry of sets and measures in Euclidean spaces. Cambridge University Press. ISBN978-0-521-65595-8.

Article Sources and Contributors

39

Article Sources and Contributors

Fractal dimension Source: http://en.wikipedia.org/w/index.php?oldid=480608392 Contributors: Ajgorhoe, Akarpe, Beaumont, Berland, Bjankuloski06en, Coemgenus, Crisfilax, Crtolle, Diego Moya, Dryazan, Eequor, Farhadbordbar, Gaius Cornelius, Gandalf61, Giftlite, Gsrdzl, Gulmammad, Headbomb, Henrikholm, Huttarl, Ideal gas equation, Igitur, Jheald, Jrolston, KDesk, KamiGolchin, Kieff, Lambiam, Leevanjackson, Lerdsuwa, Lgallindo, Llorenzi, Lpet11, MZMcBride, MarSch, Mcld, Michael Hardy, Nakon, Nazlfrag, Nico2panama, Nousernamesleft, Oshanker, Peteymills, Prokofiev2, RDBrown, Risk one, Rjwilmsi, Schneelocke, Sigmundur, Tcamps42, Tdtsi, Thyamu, Twbaroberts, Twisp, Unara, Vairis, Vcelloho, Venny85, W.F.Galway, 40 anonymous edits Fractal analysis Source: http://en.wikipedia.org/w/index.php?oldid=479434817 Contributors: Akarpe, Delaszk, Excirial, Famgroup, FatPope, Fractalmichel, Gavinthorp, Hallows AG, Hjb, Qetuth, RHaworth, Rjwilmsi, Tinucherian, 3 anonymous edits Box counting Source: http://en.wikipedia.org/w/index.php?oldid=480099360 Contributors: Akarpe, Headbomb, Michael Hardy, Niceguyedc, Questionable pulse Multifractal system Source: http://en.wikipedia.org/w/index.php?oldid=480450091 Contributors: A5b, Ael 2, Akarpe, Alansohn, Dougher, Giftlite, Kevin Baas, Kreachure, Lavateraguy, Ldecola, Linas, MFMCguy, MichaelMcGuffin, Netsnipe, RDBrown, Rjwilmsi, SimpsonDG, Stevelihn, Twbaroberts, Utopiah, 16 anonymous edits Entropy (information theory) Source: http://en.wikipedia.org/w/index.php?oldid=480571920 Contributors: Abdull, Aelkiss, Afelton, Ahoerstemeier, AlanUS, Albmont, Ale2006, Alejo2083, Alexnye, Algorithms, Alksentrs, Alpt, Andrei Stroe, Ap, Army1987, AugPi, B4hand, BD2412, Beland, Belkovich, BlackAce48, Blaisorblade, Blueyeru, Boaz, Bowsmand, Br77rino, Brandon, Brona, Bryan Derksen, Btyner, Buster79, CBM, Cassandra Cathcart, Cburnett, Cesarth73, Christopherlin, Citrus538, Coffee2theorems, CommonsDelinker, Constructive editor, Conversion script, Coppertwig, CorbinSimpson, Cretog8, Cybercobra, DMacks, Dailyknowledge, Dan Gluck, Daniel.Cardenas, Dauto, Deepmath, Derek Ross, Diegotorquemada, Djr32, Dmh, DmitriyV, Dougher, Dragon 280, Dysprosia, ESkog, Eastereaster, EdJohnston, Edchi, Elsehow, EnOreg, Entropeter, Erianna, Erkcan, FarzanehSarafraz, Fibonacci, FilipeS, FilippoSidoti, First Harmonic, Flammifer, Flashmorbid, Foobaz, Fryed-peach, Fulldecent, GEBStgo, Giftlite, Graham87, Gubbubu, Guettarda, Gusshoekey, Gutworth, Hagedis, Hakeem.gadi, Hans de Vries, Hanspi, Heysan, Hirstormandy, Husond, Ignoramibus, Informationtheory, InverseHypercube, IstvanWolf, JaGa, Jabowery, Jann.poppinga, Jdthood, Jeffq, Jetekus, Jheald, Johnuniq, JoseREMY, Jough, Kace7, Karol Langner, Kencf0618, Kestasjk, Kimchi.sg, Kjells, Kku, Kurykh, Kymacpherson, LOL, Landon1980, Linas, Lotje, Magioladitis, Male1979, MarkSweep, Maxlittle2007, Mcld, Mcstrother, Mdsam2, Mduteil, Melcombe, Mfeadler, Mhadi.afrasiabi, Michael C Price, Michael Hardy, Michael Rogers, Michel.machado, Mintrick, MisterSheik, Mitch.mcquoid, Mkweise, Music Sorter, Musides, Nbarth, Nearfar, Neilc, Neonleonb, Netpilot43556, Ninjagecko, Nneonneo, Noeckel, Nonsuch, Ohnoitsjamie, Oleg Alexandrov, Olivier, Omegatron, Orubt, OverlordQ, PAR, Paul-L, Pax:Vobiscum, Phancy Physicist, Phy1729, PhysPhD, PierreAbbat, Pmagrass, Purplie, Qwfp, R'n'B, Ra2007, Radagast83, Rade Kutil, Rc3002, Rich Farmbrough, Rick.G, Rinconsoleao, Rjwilmsi, Robma, Romanpoet, Rursus, Ryan Reich, Ryan256, Saigyo, Sbwoodside, Schuermann, Sctfn, Seanmadsen, Seaphoto, Severoon, Shaul1, Shockem, Shreevatsa, Sigma0 1, Sj, Sligocki, Snowgrouse, Snoyes, Sobec, Spakin, Spinningspark, SteveMcCluskey, Stevertigo, Stirling Newberry, SudoGhost, Svick, Swpb, TedDunning, The Anome, TheObtuseAngleOfDoom, Thomas Arelatensis, Tide rolls, TimProof, Time3000, Tkircher, Tobias Bergemann, Tobias Hoevekamp, Tomash, Tomchiukc, Tommy Herbert, Trachten, Trevorgoodchild, Tschijnmotschau, Vql, Waveguy, WikiC, Wikomidia, Wiml, Ww, Youandme, Zandr4, Zeman, Ziddy, Zylorian, 298 anonymous edits Rnyi entropy Source: http://en.wikipedia.org/w/index.php?oldid=480714131 Contributors: Aetheling, Bailo26, Calbaer, Charles Matthews, Clovis Sangrail, DRLB, David Eppstein, Entropeter, Gene Nygaard, HappyCamper, Headbomb, Hexiang, JASAlways, Jheald, Korepin, Ligulem, Linas, Michael Hardy, Moala, Ohjaek33, Rjwilmsi, Schmei, Shakir, Vegalabs, Wullj, Ywaz, 28 anonymous edits Hausdorff dimension Source: http://en.wikipedia.org/w/index.php?oldid=474428036 Contributors: Aleph4, Alex Schreiber, AugPi, AxelBoldt, BernardH, BigK101, Bomazi, Bonev, CRGreathouse, CSTAR, Charles Matthews, Chebyshev, Cheeser1, CiaPan, Ciphergoth2, ComplexZeta, Conversion script, David Eppstein, DavidCary, Dcoetzee, Diannaa, Dirac1933, Drstuey, Druiffic, Eequor, Everyking, Expert in topology, Feraudyh, Furrykef, Gadykozma, Gdr, Gene Nygaard, George Burgess, Gervasecb, Giftlite, Graham87, Graphite Elbow, Habj, HighwoodFool, IVAN3MAN, Iacommo, JF Manning, Jannex, Joriki, JunCTionS, Kaini, Kakofonous, Karshon, Kdammers, Kier07, Koeplinger, Lethe, Linas, Link hyrule5, MIT Trekkie, MathMartin, Meocjt, Michael Hardy, Mohsen.soltanifar, Mpd1989, Naraht, Nol Aders, OdedSchramm, Oleg Alexandrov, PL290, Pjacobi, Poccil, Poor Yorick, Prokofiev2, RDBrown, Rgrg, Rjwilmsi, Rlove, Sade, Sam Hocevar, Schmloof, Schneelocke, Scleria, Scorwin, Slightsmile, Smjg, Solkoll, Somejan, TeeEmCee, The Anome, TheSeven, Thenub314, Tinus, Tobias Bergemann, TomyDuby, Topology Expert, Tosha, TurilCronburg, Uffish, User23456789, Volland, WLior, Wikimol, WingedPig, Wood Thrush, XJamRastafire, Zeromaru, Zundark, 88 anonymous edits

Image Sources, Licenses and Contributors

40

Image Sources, Licenses and Contributors

Image:britain-fractal-coastline-200km.png Source: http://en.wikipedia.org/w/index.php?title=File:Britain-fractal-coastline-200km.png License: GNU Free Documentation License Contributors: User:Acadac, User:Avsa, User:Wapcaplet Image:britain-fractal-coastline-100km.png Source: http://en.wikipedia.org/w/index.php?title=File:Britain-fractal-coastline-100km.png License: GNU Free Documentation License Contributors: User:Acadac, User:Avsa, User:Wapcaplet Image:britain-fractal-coastline-50km.png Source: http://en.wikipedia.org/w/index.php?title=File:Britain-fractal-coastline-50km.png License: GNU Free Documentation License Contributors: User:Acadac, User:Avsa, User:Wapcaplet Image:32 segment fractal.jpg Source: http://en.wikipedia.org/w/index.php?title=File:32_segment_fractal.jpg License: Creative Commons Attribution-Sharealike 3.0 Contributors: User:Akarpe Image:blueklineani2.gif Source: http://en.wikipedia.org/w/index.php?title=File:Blueklineani2.gif License: Creative Commons Attribution-Sharealike 3.0 Contributors: User:Akarpe Image:Fractaldimensionexample.PNG Source: http://en.wikipedia.org/w/index.php?title=File:Fractaldimensionexample.PNG License: Public Domain Contributors: Brendan Ryan. Original uploader was Nazlfrag at en.wikipedia Image:KochFlake.svg Source: http://en.wikipedia.org/w/index.php?title=File:KochFlake.svg License: GNU Free Documentation License Contributors: D-Kuru, Wxs file:onetwosix.png Source: http://en.wikipedia.org/w/index.php?title=File:Onetwosix.png License: Creative Commons Attribution-Sharealike 3.0 Contributors: User:Akarpe Image:Binarizing neuron image.gif Source: http://en.wikipedia.org/w/index.php?title=File:Binarizing_neuron_image.gif License: Creative Commons Attribution-Sharealike 3.0 Contributors: User:Akarpe Image:Slidestack.gif Source: http://en.wikipedia.org/w/index.php?title=File:Slidestack.gif License: Creative Commons Attribution-Sharealike 3.0 Contributors: User:Akarpe Image:lcfd.gif Source: http://en.wikipedia.org/w/index.php?title=File:Lcfd.gif License: Creative Commons Attribution-Sharealike 3.0 Contributors: User:Akarpe Image:Fixedstack.gif Source: http://en.wikipedia.org/w/index.php?title=File:Fixedstack.gif License: Creative Commons Attribution-Sharealike 3.0 Contributors: User:Akarpe Image:Retina lcfd.gif Source: http://en.wikipedia.org/w/index.php?title=File:Retina_lcfd.gif License: Creative Commons Attribution-Sharealike 3.0 Contributors: User:Akarpe Image:optimal covering grids.png Source: http://en.wikipedia.org/w/index.php?title=File:Optimal_covering_grids.png License: Creative Commons Attribution-Sharealike 3.0 Contributors: User:Akarpe Image:Karperien Strange Attractor 200.gif Source: http://en.wikipedia.org/w/index.php?title=File:Karperien_Strange_Attractor_200.gif License: Creative Commons Attribution-Sharealike 3.0 Contributors: User:Akarpe File:Distort.gif Source: http://en.wikipedia.org/w/index.php?title=File:Distort.gif License: Creative Commons Attribution-Sharealike 3.0 Contributors: User:Akarpe File:Dqvsq.gif Source: http://en.wikipedia.org/w/index.php?title=File:Dqvsq.gif License: Creative Commons Attribution-Sharealike 3.0 Contributors: User:Akarpe File:Binary entropy plot.svg Source: http://en.wikipedia.org/w/index.php?title=File:Binary_entropy_plot.svg License: GNU Free Documentation License Contributors: Brona and Alessio Damato image:Great Britain Hausdorff.svg Source: http://en.wikipedia.org/w/index.php?title=File:Great_Britain_Hausdorff.svg License: Creative Commons Attribution-Sharealike 3.0 Contributors: Prokofiev Image:Sierpinski deep.svg Source: http://en.wikipedia.org/w/index.php?title=File:Sierpinski_deep.svg License: Creative Commons Attribution-Sharealike 3.0 Contributors: Sega sai

License

41

License
Creative Commons Attribution-Share Alike 3.0 Unported //creativecommons.org/licenses/by-sa/3.0/