Scribd
Upload
51 views12 pages

An Introduction To Kohonen Self Organizing Maps: Rajarshi Guha

This document is an introduction to Kohonen self-organizing maps (SOMs). SOMs are a type of competitive learning neural network that maps high-dimensional input onto a low-dimensional grid. They are useful for visualizing and detecting patterns in multidimensional data. The SOM training process involves presenting input vectors to the grid and modifying the best matching and neighboring units to better match the input. This leads to an ordered mapping with similar inputs mapped to nearby regions of the grid.

Uploaded by

Karan
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
51 views12 pages

An Introduction To Kohonen Self Organizing Maps: Rajarshi Guha

This document is an introduction to Kohonen self-organizing maps (SOMs). SOMs are a type of competitive learning neural network that maps high-dimensional input onto a low-dimensional grid. They are useful for visualizing and detecting patterns in multidimensional data. The SOM training process involves presenting input vectors to the grid and modifying the best matching and neighboring units to better match the input. This leads to an ordered mapping with similar inputs mapped to nearby regions of the grid.

Uploaded by

Karan
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 12

www.Vidyarthiplus.

com

An Introduction to Kohonen Self


Organizing Maps
Rajarshi Guha

Penn State University

An Introduction to Kohonen Self Organizing Maps – p.1/12


www.Vidyarthiplus.com

Kohonen Self Organizing


Maps
What are SOM’s ?
Competitive Learning ANN
Elastic net of points which is made to fit the
input vector

Mode of Functioning
Each unit of the map recieves identical
inputs
The units compete for selection
The selected neuron and surrounding
neighbours get modified An Introduction to Kohonen Self Organizing Maps – p.2/12
www.Vidyarthiplus.com

Kohonen Self Organizing


Maps
Utility
Visualizing N dimensional data in 2D
Detecing similarity and degree’s of
similarity

An Introduction to Kohonen Self Organizing Maps – p.3/12


www.Vidyarthiplus.com

Structure of the Map


Square grid
Each grid point is a
vector containing the
descriptor values
The grid wraps round
the edges
The grid is initialized
with random vectors

An Introduction to Kohonen Self Organizing Maps – p.4/12


www.Vidyarthiplus.com

Training the Map


Each descriptor vector in the training set is
presented to all grid points.
Select the closest matching grid point based
on minimum Euclidean distance
v
um
uX
dsj = t (s si − wji )
i=1

Modify the selected grid point and its


neighbours.
Degree of modification reduces with each
training iteration An Introduction to Kohonen Self Organizing Maps – p.5/12
www.Vidyarthiplus.com

Modification of Grid
Points, aka Learning
Selected grid point and topologically close
gridpoints are modified
Leads to smoothing which leads to global
ordering
The general form of modification is
mi (t + 1) = mi (t) + hci (t)[x(t) − mi (t)]
hci (t) is termed as the neighborhood function
As t → ∞, hci (t) → 0

An Introduction to Kohonen Self Organizing Maps – p.6/12


www.Vidyarthiplus.com

The Neighborhood
Function
The neighborhood function can be

of various forms.

Neighborhood Set
hci (t) = α(t) if An Introduction to Kohonen Self Organizing Maps – p.7/12
www.Vidyarthiplus.com

The Neighborhood
Function
Gaussian Neighborhood Function
 2

krc −ri k
hci (t) = α(t) exp − 2σ2 (t)
σ is the width of the smoothening
Corresponds to the radius of the
neighborhood set

An Introduction to Kohonen Self Organizing Maps – p.8/12


www.Vidyarthiplus.com

The Learning Factor


α(t) is termed the learning factor. Some
properties of the function are:
0<α<1
Decreases monotonically with time
When α = 0, learning stops
α can be varied in several ways
Constant decrement
Function of training iteration
Recursive - αi (t + 1) = αi (t)/(1 + hci αi (t))
An Introduction to Kohonen Self Organizing Maps – p.9/12
www.Vidyarthiplus.com

Classification
Assign an arbitrary class to the first grid point
For each grid point calculate distances to
each nearest neighbor
If the distance to a neighbor is less than a
user specified threshold then the neighbor is
in the same class as the grid point
Finally, find the closest matching neuron in
the training set for the initial grid point.
Correspondingly modify class assignments of
all the other grid points
An Introduction to Kohonen Self Organizing Maps – p.10/12
www.Vidyarthiplus.com

More to Come

An Introduction to Kohonen Self Organizing Maps – p.11/12


www.Vidyarthiplus.com

References
[1] Lingran, C.; Gasteiger, J. J. Am. Chem. Soc. 1997, 119, 4033.
[2] Gasteiger, J.; Zupan, J. Angew. Chem. Intl. Ed. Engl. 1993, 32, 503.
[3] Kohonen, T. Self Organizing Maps; Springer Series in Information Sciences
Springer: Espoo, Finland, 1994.

An Introduction to Kohonen Self Organizing Maps – p.12/12

You might also like