Yours Truly

Sunil Template
Contents

List of Figures ix

List of Tables xi

I This is What a Part Would Look Like 1

1 Basic Concepts 3

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1.1 A component part . . . . . . . . . . . . . . . . . . . . 3
1.2 Record Linkage Model . . . . . . . . . . . . . . . . . . . . . . 7
1.2.1 A component part . . . . . . . . . . . . . . . . . . . . 7
1.2.1.1 H3 A component part . . . . . . . . . . . . 8
1.3 Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

Bibliography 13

vii
List of Figures

1.1 List of gure caption goes here . . . . . . . . . . . . . . . . 7

1.2 Short gure caption . . . . . . . . . . . . . . . . . . . . . . . 8
1.3 The bar charts depict the dierent risk contributions . . . . 9

ix
List of Tables

( )
1.1 Now we are engaged (aag ) aag in a great civil war, testing
whether that nation, or any nation so conceived. . . . . . . 4
1.2 Examples for illustrating attacks . . . . . . . . . . . . . . . 7

xi
Symbol Description
To solve the generator main- annealing and genetic algo-
tenance scheduling, in the rithms have also been tested.
past, several mathematical abc This paper presents a survey
techniques have been ap- of the literature
plied. over the past fteen years in
2 These include integer pro- the generator
gramming, integer linear maintenance scheduling.
programming, dynamic pro- The objective is to
gramming, branch and sdf present a clear picture of the
bound etc. available recent literature
Several heuristic search al- ewq of the problem, the con-
gorithms have also been de- straints and the other as-
veloped. In recent years ex- pects of
pert systems, bvcn the generator maintenance
abc fuzzy approaches, simulated schedule.
 Part I

This is What a Part Would

Look Like

1
Chapter 1
Basic Concepts

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1.1 A component part . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Record Linkage Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.2.1 A component part . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.2.1.1 H3 A component part . . . . . . . . . . . . . . . . . . . . . 8
1.3 Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

1.1 Introduction
The term reliability usually refers to the probability that a component
or system will operate satisfactorily either at any particular instant at which
it is required or for a certain length of time. Fundamental to quantifying
reliability s a knowledge of how to dene, assess and combine probabilities [1].
This may hinge on identifying the form of the variability which is nherent n
most processes. If all components had a xed known lifetime there would be
no need to model reliability.

1.1.1 A component part

A component part for an electronic item is manufactured at one of three
dierent factories, and then delivered to the main assembly line.Of the total
number supplied, factory A supplies 50%, factory B 30%, and factory C 20%.
Of the components manufactured at factory A, 1% are faulty and the cor-
responding proportions for factories B and C are 4% and 2% respectively. A
component is picked at random from the assembly line. What is the probability
that it is faulty [4]? A component part for an electronic item is manufactured
at one of three dierent factories, and then delivered to the main assembly
line.Of the total number supplied, factory A supplies 50%, factory B 30%,
and factory C 20%. Of the components manufactured at factory A, 1% are
faulty and the corresponding proportions for factories B and C are 4% and 2%
respectively. A component is picked at random from the assembly line. What
is the probability that it is faulty? A component part for an electronic item
is manufactured at one of three dierent factories, and then delivered to the
main assembly line.Of the total number supplied, factory. A component part

3
4 Sunil Template
TABLE 1.1: Now we are engaged (aag )
( )
aag in a great civil war, testing whether that
nation, or any nation so conceived.
Scene Reg. fts. Hor. fts. Ver. fts.
Ball 19, 221 4, 598 3, 200
Pepsia 46, 281 6, 898 5, 400
Keybrdb 27, 290 2, 968 3, 405
Pepsi 14, 796 9, 188 3, 209

for an electronic item is manufactured at one of three dierent factories, and

then delivered to the main assembly line.Of the total number supplied, fac-
tory A supplies 50%, factory B 30%, and factory C 20%. Of the components
manufactured at factory A, 1% are faulty and the corresponding proportions
for factories B and C are 4% and 2% respectively. A component is picked at
random from the assembly line. What is the probability that it is faulty?

A Process is a structured, measured set of activities de-

signed to produce a specic output for a particular customer
or marketA process is thus a specic ordering of work ac-
tivities across time and space, with a beginning, an end. and
clearly dened inputs and outputs: a structure for action.

Thomas Davenport
Senior Adjutant to the Junior Marketing VP

MultiRelational k-Anonymity. Most works on k-anonymity focus on

anonymizing a single data table; however, a real-life [2] database usually con-
tains multiple relational tables. This has proposed a privacy model called Mul-
tiR k-anonymity to ensure k-anonymity on multiple relational tables. Their
model assumes that a relational database contains a person-specic table P T
and a set of tables T1 , , Tn , where P T contains a person identier P id and
some sensitive attributes, and Ti , for 1 i n, contains some foreign keys,
some attributes in QID, and sensitive attributes. The general privacy notion
is to ensure that for each record owner o contained in the join of all tables
PT o n T1 on on Tn , there exists at least k 1 other record owners share
the same QID with o. It is important to emphasize that the k-anonymization
is applied at the record owner level, not at the record level in traditional k-
anonymity. This idea is similar to (X, Y )-anonymity, where X = QID and
Y = {P id}.
A component part for an electronic item is [3] manufactured at one of
three dierent factories, and then delivered to the main assembly line.Of the
 Basic Concepts 5

total number supplied, factory A supplies 50%, factory B 30%, and factory C
20%. Of the components manufactured at factory A, 1% are faulty and the
corresponding proportions for factories B and C are 4% and 2% respectively. A
component is picked at random from the assembly line. What is the probability
that it is faulty?
In most literature on PPDP, they [6] consider a more relaxed, yet more
practical, notion of privacy protection by assuming limited attackers back-
ground knowledge. Below, the term victim refers to the record owner being
linked. We can broadly classify linking models to two families.

A component part for an electronic item is manufactured at one of

three dierent factories, and then delivered to the main assembly line.Of
the total number supplied, factory A supplies 50%, factory B 30%, and
factory C 20%. Of the components manufactured at factory A, 1% are
faulty and the corresponding proportions for factories B and C are 4%
and 2% respectively. A component is picked at random from the assembly
line. What is the probability that it is faulty?

A component part for an electronic item is manufactured at one of three

dierent factories, and then delivered to the main assembly line.Of the total
number supplied, factory A supplies 50%, factory B 30%, and factory C 20%.
Of the components manufactured at factory A, 1% are faulty and the corre-
sponding proportions for factories B and C are 4% and 2% respectively. A
component is picked at random from the assembly line. What is the probabil-
ity that it is faulty?
One family considers a privacy threat occurs when an attacker is able
to link a record owner to a record in a published data table, to a sensitive
attribute in a published data table, or to the published data table itself. We
call them record linkage, attribute linkage, and table linkage, respectively. In all
types of linkages, we assume that the attacker knows the QID of the victim.
In record and attribute linkages, we further assume that the attacker knows
the presence of the victims record in the released table, and seeks to identify
the victims record and/or sensitive information from the table [10]. In table
linkage, the attack seeks to determine the present or absent of the victims
record in the released table. A data table is considered to privacy preserved
if the table can eectively prevent the attacker from successfully performing
these types of linkages on the table [7]. Sections 1.1-1.2 study this family of
privacy models.

t
t
t
t
b =
var var (j k ) = 2 (1/nj + 1/nk ). (1.1)
j=1 k=j+1 j=1 k=j+1
6 Sunil Template

An obvious measure of imbalance is just the dierence in the number of

times the two treatments are allocated

Dn = |nA nB |. (1.2)

For rules such as deterministic allocation, for which the expected value of this
dierence can be calculated, we obtain the population value Dn .

Box Title Here

Another family aims at achieving the uninformative principle: The pub-
lished table should provide the attacker with little additional information
beyond the background knowledge. There should not be a large dier-
ence between the prior and posterior beliefs; otherwise, there is a privacy
threat [5, 6]. Many privacy models in this family are designed for statisti-
cal database and do not distinguish attributes in T into QID, but some
of them could also thwart record, attribute, and table linkages. Section 1.1
studies this family of privacy models.
Let m be a prime number. With the addition and multiplication as
dened above, Zm is a eld.

Another family aims at achieving the uninformative principle: The pub-

lished table should provide the attacker with little additional information be-
yond the background knowledge. There should not be a large dierence be-
tween the prior and posterior beliefs; otherwise, there is a privacy threat [5, 6].
Many privacy models in this family are designed for statistical database and
do not distinguish attributes in T into QID, but some of them could also
thwart record, attribute, and table linkages. Section 1.1 studies this family of
privacy models.

Theorem 1 Let m be a prime number. With the addition and multiplication

as defined above, Zm is a field.

Proof 1 Most of the proof of this theorem is routine. It is clear that 0 Zm

and 1 Zm are the zero element and identity element. If a Zm and a = 0,
then m a is the additive inverse of a. If a Zm and a = 0, then the greatest
common divisor of a and m is 1, and hence there exist integers s and t such
that sa + tm = 1. Thus sa = 1 tm is congruent to 1 modulo m. Let s be the
integer in Zm congruent to s modulo m. Then we also have s a 1 mod m.
Hence s is the multiplicative inverse of a modulo m. Verification of the rest
of the field properties is now routine.

Note: The process of integrating reengineering is best accomplished with an

engineer, a dog, and a cat.
 Basic Concepts 7
TABLE 1.2: Examples for
illustrating attacks
Job Sex Age Disease
Engineer Male 35 Hepatitis
Engineer Male 38 Hepatitis
Lawyer Male 38 HIV
Writer Female 30 Flu
Writer Female 30 HIV
Dancer Female 30 HIV
Dancer Female 30 HIV

1.2 Record Linkage Model

In the privacy attack of record linkage, some value qid on QID identies a
small number of records in the released table T , called a group. If the victims
QID matches the value qid, the victim is vulnerable to being linked to the
small number of records in the group [8]. In this case, the attacker faces only
a small number of possibilities for the victims record, and with the help
of additional knowledge, there is a chance that the attacker could uniquely
identify the victims record from the group.

1.2.1 A component part

A component part for an electronic item is manufactured at one of three
dierent factories, and then delivered to the main assembly line.Of the total
number supplied, factory A supplies 50%, factory B 30%, and factory C 20%.

FIGURE 1.1: Figure caption goes here. Figure caption goes here.
8 Sunil Template

Of the components manufactured at factory A, 1% are faulty and the corre-

sponding proportions for factories B and C are 4% and 2% respectively. A
component is picked at random from the assembly line. What is the probabil-
ity that it is faulty?

FIGURE 1.2: Figure caption goes here. Figure caption goes here. Figure
caption goes here.

1.2.1.1 H3 A component part

A component part for an electronic item is manufactured at one of three
[9] dierent factories, and then delivered to the main assembly line.Of the
total number supplied, factory A supplies 50%, factory B 30%, and factory C
20%. Of the components manufactured at factory A, 1% are faulty and the
corresponding proportions for factories B and C are 4% and 2% respectively. A
component is picked at random from the assembly line. What is the probability
that it is faulty?
A fundamental notion [10] is that of a subspace of F n . Let V be a nonempty
 Basic Concepts 9

4
x 10
4.5

3.5

3
frequency

2.5

1.5

0.5

0
0 0.05 0.1 0.15
loss in % of exposure
(a)
4
x 10
4.5

3.5

3
frequency

2.5

1.5

0.5

0
0 0.05 0.1 0.15
loss in % of exposure
(b)

FIGURE 1.3: The bar charts depict the dierent risk contributions (top:
99% quantile, bottom: 99.9% quantile) of the business areas of a bank. The
black bars are based on a Var/Covar approach, the white ones correspond to
shortfall risk.
10 Sunil Template

subset of F n . Then V is a subspace of F n provided V is closed under vector

addition and scalar multiplication, that is,

(a) For all u and v in V , u + v is also in V .

(b) For all u in V and c in F , cu is in V .

Let u be in the subspace V . Because 0u = 0, it follows that the zero vector

is in V . Similarly, u is in V for all u in V . A simple example of a subspace
of F n is the set of all vectors (0, a2 , . . . , an ) with rst coordinate equal to 0.
The zero vector itself is a subspace.

Think About It...

Commonly thought of as the rst modern computer, EN-
TAC was built in 1944. It took up more space than an 18-
wheelers tractor trailer and weighed more than 17 Chevrolet
Camaros. It consumed 140,000 watts of electricity while ex-
ecuting up to 5,000 basic arithmetic operations per second.
One of todays popular microprocessors, the 486, is built on
a tiny piece of silicon about the size of a dime.
With the continual expansion of capabilities, computing
power will eventually exceed the capacity for human com-
prehension or human control.

The Information Revolution

Business Week

Definition 1 Let u(1) , u(2) , . . . , u(m) be vectors in F n , and let c1 , c2 , . . . , cm

be scalars. Then the vector

c1 u(1) + c2 u(2) + + cm u(m)

is called a linear combination of u(1) , u(2) , . . . , u(m) . If V is a subspace of

F n , then V is closed under vector addition and scalar multiplication, and
it follows easily by induction that a linear combination of vectors in V is
also a vector in V . Thus subspaces are closed under linear combinations; in
fact, this can be taken as the dening property of subspaces. The vectors
u(1) , u(2) , . . . , u(m) span V (equivalently, form a spanning set of V ) provided
every vector in V is a linear combination of u(1) , u(2) , . . . , u(m) . The zero vector
can be written as a linear combination of u(1) , u(2) , . . . , u(m) with all scalars
equal to 0; this is a trivial linear combination. The vectors u(1) , u(2) , . . . , u(m)
 Basic Concepts 11

are linearly dependent provided there are scalars c1 , c2 , . . . , cm , not all of which
are zero, such that

c1 u(1) + c2 u(2) + + cm u(m) = 0,

that is, the zero vector can be written as a nontrivial linear combination of
u(1) , u(2) , . . . , u(m) . For example, the vectors (1, 4), (3, 1), and (3, 5) in 2
are linearly dependent since

3(1, 4) + 1(3, 2) 2(3, 5) = (0, 0).

Vectors are linearly independent provided they are not linearly dependent.
The vectors u(1) , u(2) , . . . , u(m) are a basis of V provided they are linearly
independent and span V . By an ordered basis we mean a basis in which the
vectors of the basis are listed in a specied order; to indicate that we have
an ordered basis we write (u(1) , u(2) , . . . , u(m) ). A spanning set S of V is a
minimal spanning set of V provided that each set of vectors obtained from S
by removing a vector is not a spanning set for V . A linearly independent set S
of vectors of V is a maximal linearly independent set of vectors of V provided
that for each vector w of V that is not in S, S {w} is linearly dependent
(when this happens, w must be a linear combination of the vectors in S). 

In addition to matrix addition, subtraction, and multiplication, there is one

additional operation that we dene now. Its perhaps the simplest of them all.
Let A = [aij ] be an m by n matrix and let c be a number [3]. Then the matrix
c A, or simply cA, is the m by n matrix obtained by multiplying each entry
of A by c:
cA = [caij ].
The matrix cA is called a scalar multiple of A.

A component part for an electronic item is [3] manufactured at one

of three dierent factories, and then delivered to the main assembly
line.Of the total number supplied, factory A supplies 50%, factory B
30%, and factory C 20%. Of the components manufactured at factory
A, 1% are faulty and the corresponding proportions for factories B
and C are 4% and 2% respectively.

Commonly thought of as the rst modern computer, ENTAC was built in

1944. It took up more space than an 18-wheelers tractor trailer and weighed
more than 17 Chevrolet Camaros. It consumed 140,000 watts of electricity
while executing up to 5,000 basic arithmetic operations per second. One of
todays popular microprocessors, the 486, is built on a tiny piece of silicon
about the size of a dime.
12 Sunil Template

1.3 Glossary
360 Degree Review: Performance review that includes feedback from su-
periors, peers, subordinates, and clients.

Abnormal Variation: Changes in process performance that cannot be ac-

counted for by typical day-to-day variation. Also referred to as non-
random variation.
Acceptable Quality Level (AQL): The minimum number of parts that
must comply with quality standards, usually stated as a percentage.
Activity: The tasks performed to change inputs into outputs.

Adaptable: An adaptable process is designed to maintain eectiveness and

eciency as requirements change. The process is deemed adaptable when
there is agreement among suppliers, owners, and customers that the pro-
cess will meet requirements throughout the strategic period.
Bibliography

[1] G. Bontempi and Y. Le Borgne. An adaptive modular approach to the

mining of sensor network data. In Proceedings of the Workshop on Data
Mining in Sensor Networks, SIAM SDM, pages 39. SIAM Press, 2005.

[2] KI Diamantaras and SY Kung. Principal component neural networks:

theory and applications. John Wiley & Sons, Inc. New York, NY, USA,
1996.
[3] A. Hyvarinen, J. Karhunen, and E. Oja. Independent Component Anal-
ysis. J. Wiley New York, 2001.
[4] M. Ilyas, I. Mahgoub, and L. Kelly. Handbook of Sensor Networks: Com-
pact Wireless and Wired Sensing Systems. CRC Press, Inc. Boca Raton,
FL, USA, 2004.

[5] A. Jain and E.Y. Chang. Adaptive sampling for sensor networks. ACM
International Conference Proceeding Series, pages 1016, 2004.

[6] I.T. Jollie. Principal Component Analysis. Springer, 2002.

[7] S. Madden, M.J. Franklin, J.M. Hellerstein, and W. Hong. TAG: a Tiny
AGgregation Service for Ad-Hoc Sensor Networks. In Proceedings of the
5th ACM Symposium on Operating System Design and Implementation
(OSDI), volume 36, pages 131 146. ACM Press, 2002.

[8] S.R. Madden, M.J. Franklin, J.M. Hellerstein, and W. Hong. TinyDB: an
acquisitional query processing system for sensor networks. ACM Trans-
actions on Database Systems (TODS), 30(1):122173, 2005.
[9] K.V. Mardia, J.T. Kent, J.M. Bibby, et al. Multivariate analysis. Aca-
demic Press New York, 1979.
[10] Y. Yao and J. Gehrke. The cougar approach to in-network query pro-
cessing in sensor networks. ACM SIGMOD Record, 31(3):918, 2002.

13