Foundations of Data Science 1st Edition Avrim

Blum pdf download

https://ebookfinal.com/download/foundations-of-data-science-1st-
edition-avrim-blum/

Explore and download more ebooks or textbooks

at ebookfinal.com
We have selected some products that you may be interested in
Click the link to download now or visit ebookfinal.com
for more options!.

The foundations of EU data protection law 1st Edition

Lynskey

https://ebookfinal.com/download/the-foundations-of-eu-data-protection-
law-1st-edition-lynskey/

The Foundations of Science Science and Hypothesis The

Value of Science Science and Method Henri Poincaré

https://ebookfinal.com/download/the-foundations-of-science-science-
and-hypothesis-the-value-of-science-science-and-method-henri-poincare/

Isaac Asimov The Foundations of Science Fiction Gunn

https://ebookfinal.com/download/isaac-asimov-the-foundations-of-
science-fiction-gunn/

Action Science Foundations of an Emerging Discipline 1st

Edition Wolfgang Prinz

https://ebookfinal.com/download/action-science-foundations-of-an-
emerging-discipline-1st-edition-wolfgang-prinz/
Data Intensive Science 1st Edition Terence Critchlow

https://ebookfinal.com/download/data-intensive-science-1st-edition-
terence-critchlow/

Foundations of Inference in Natural Science 1st Edition

John Oulton Wisdom (Editor)

https://ebookfinal.com/download/foundations-of-inference-in-natural-
science-1st-edition-john-oulton-wisdom-editor/

Philosophers of the Renaissance 1st Edition Paul Richard

Blum

https://ebookfinal.com/download/philosophers-of-the-renaissance-1st-
edition-paul-richard-blum/

The Water Cycle Science Foundations 1st Edition Nikole

Brooks Bethea

https://ebookfinal.com/download/the-water-cycle-science-
foundations-1st-edition-nikole-brooks-bethea/

Service Science The Foundations of Service Engineering and

Management 1st Edition Robin G. Qiu

https://ebookfinal.com/download/service-science-the-foundations-of-
service-engineering-and-management-1st-edition-robin-g-qiu/
Foundations of Data Science 1st Edition Avrim Blum
Digital Instant Download
Author(s): Avrim Blum, John Hopcroft, Ravindran Kannan
ISBN(s): 9781108485067, 1108485065
Edition: 1
File Details: PDF, 10.73 MB
Year: 2020
Language: english
 Foundations of Data Science 1

John Hopcroft

Ravindran Kannan

Version 21/8/2014

These notes are a rst draft of a book being written by Hopcroft and Kannan and
in many places are incomplete. However, the notes are in good enough shape to prepare
lectures for a modern theoretical course in computer science. Please do not put solutions
to exercises online as it is important for students to work out solutions for themselves
rather than copy them from the internet.

Thanks
JEH

1Copyright 2011. All rights reserved

1
Contents
1 Introduction 7

2 High-Dimensional Space 10
2.1 Properties of High-Dimensional Space ................................................................... 12
2.2 The Law of Large Numbers .................................................................................... 13
2.3 The High-Dimensional Sphere ................................................................................. 15
2.3.1 TheSphere and the Cube in High Dimensions ........................................ 16
2.3.2 Volume and Surface Area of the Unit Sphere ........................................ 17
2.3.3 The Volume is Near the Equator ............................................................ 20
2.3.4 The Volume is in a Narrow Annulus ......................................................... 23
2.3.5 The Surface Area is Near the Equator .................................................. 24
2.4 Volumes of Other Solids ........................................................................................... 26
2.5 Generating Points Uniformly at Random on the Surfaceof a Sphere ............ 27
2.6 Gaussians in High Dimension ................................................................................. 27
2.7 Bounds on Tail Probability ........................................................................................ 33
2.8 Applications of the tail bound ................................................................................. 35
2.9 Random Projection and Johnson-LindenstraussTheorem ................................ 38
2.10 Bibliographic Notes ..................................................................................................... 41
2.11 Exercises ......................................................................................................................... 42

3 Best-Fit Subspaces and Singular Value Decomposition (SVD) 52

3.1 Singular Vectors ............................................................................................................ 53
3.2 Singular Value Decomposition (SVD) .................................................................... 56
3.3 Best Rank k Approximations ................................................................................. 58
3.4 Left Singular Vectors .................................................................................................. 60
3.5 Power Method for Computing the Singular Value Decomposition ................. 62
3.6 Applications of Singular Value Decomposition ................................................... 64
3.6.1 Principal Component Analysis 64
3.6.2 Clustering a Mixture of Spherical Gaussians ........................................ 65
3.6.3 Spectral Decomposition 71
3.6.4 Singular Vectors and Ranking Documents ............................................ 71
3.6.5 An Application of SVD to a Discrete Optimization Problem .......... 72
3.7 Singular Vectors and Eigenvectors .......................................................................... 75
3.8 Bibliographic Notes ..................................................................................................... 76
3.9 Exercises ......................................................................................................................... 77

4 Random Graphs 85
4.1 The G(n; p) Model ..................................................................................................... 85
4.1.1 Degree Distribution ....................................................................................... 86
4.1.2 Existence of Triangles in G(n; d=n) 91
4.2 Phase Transitions ........................................................................................................ 93
4.3 The Giant Component .............................................................................................. 101

2
 4.4 Branching Processes ................................................................................................. 109
4.5 Cycles and Full Connectivity .................................................................................... 115
4.5.1 Emergence of Cycles .................................................................................... 115
4.5.2 Full Connectivity ........................................................................................... 116
4.5.3 Threshold for O(ln n) Diameter ................................................................ 118
4.6 Phase Transitions for Increasing Properties .......................................................... 119
4.7 Phase Transitions for CNF-sat ................................................................................. 121
4.8 Nonuniform and Growth Modelsof Random Graphs .......................................... 126
4.8.1 Nonuniform Models ....................................................................................... 126
4.8.2 Giant Component in Random Graphs with Given Degree Distribution127
4.9 Growth Models ........................................................................................................... 128
4.9.1 Growth Model Without Preferential Attachment 128
4.9.2 Growth Model With Preferential Attachment 135
4.10 Small World Graphs ................................................................................................ 136
4.11 Bibliographic Notes .................................................................................................... 141
4.12 Exercises ........................................................................................................................ 142

5 Random Walks and Markov Chains 153

5.1 Stationary Distribution .............................................................................................. 156
5.2 Electrical Networks and Random Walks ............................................................... 158
5.3 Random Walks on Undirected Graphs with Unit Edge Weights .................... 162
5.4 Random Walks in Euclidean Space ...................................................................... 169
5.5 The Web as a Markov Chain .................................................................................... 173
5.6 Markov Chain Monte Carlo .................................................................................... 177
5.6.1 Metropolis-HastingAlgorithm ..................................................................... 178
5.6.2 Gibbs Sampling .............................................................................................. 180
5.7 Areas and Volumes .................................................................................................... 182
5.8 Convergence of Random Walks on Undirected Graphs ..................................... 183
5.8.1 Using Normalized Conductance to Prove Convergence ....................... 188
5.9 Bibliographic Notes .................................................................................................... 191
5.10 Exercises ........................................................................................................................ 192

6 Learning and VC-dimension 202

6.1 Learning ......................................................................................................................... 202
6.2 Linear Separators, the Perceptron Algorithm, and Margins ........................... 204
6.3 Nonlinear Separators, Support Vector Machines, and Kernels ....................... 209
6.4 Strong and Weak Learning - Boosting ................................................................... 214
6.5 Number of Examples Needed for Prediction: VC-Dimension ....................... 216
6.6 Vapnik-Chervonenkis or VC-Dimension ............................................................... 219
6.6.1 Examples of Set Systems and Their VC-Dimension ........................... 220
6.6.2 The Shatter Function 223
6.6.3 Shatter Function for Set Systems of Bounded VC-Dimension . . . 224
6.6.4 Intersection Systems 226

3
 6.7 The VC Theorem ........................................................................................................ 226
6.8 Simple Learning ........................................................................................................... 229
6.9 Bibliographic Notes .................................................................................................... 230
6.10 Exercises ........................................................................................................................ 231

7 Algorithms for Massive Data Problems 238

7.1 Frequency Moments of Data Streams ................................................................... 238
7.1.1 Number of Distinct Elements in a Data Stream .................................. 239
7.1.2 Counting the Number of Occurrences of aGiven Element. ................ 243
7.1.3 Counting Frequent Elements ....................................................................... 243
7.1.4 The Second Moment ..................................................................................... 245
7.2 Matrix Algorithms Using Sampling ...................................................................... 248
7.2.1 Matrix Multiplication Using Sampling ................................................... 248
7.2.2 Sketch of a Large Matrix .............................................................................. 250
7.3 Sketches of Documents ............................................................................................ 253
7.4 Exercises ......................................................................................................................... 256

8 Clustering 260
8.1 Some Clustering Examples ..................................................................................... 260
8.2 A k-means Clustering Algorithm .......................................................................... 263
8.3 A Greedy Algorithm for k-Center Criterion Clustering ................................. 265
8.4 Spectral Clustering ..................................................................................................... 266
8.5 Recursive Clustering Based on Sparse Cuts ......................................................... 273
8.6 Kernel Methods ........................................................................................................... 274
8.7 Agglomerative Clustering .......................................................................................... 276
8.8 Dense Submatrices and Communities ................................................................... 278
8.9 Flow Methods ............................................................................................................... 281
8.10 Finding a Local Cluster Without Examining the Whole Graph ................... 284
8.11 Axioms for Clustering ............................................................................................ 289
8.11.1 An ImpossibilityResult ............................................................................... 289
8.11.2 A Satisable Set ofAxioms ........................................................................ 295
8.12 Exercises ........................................................................................................................ 297

9 Topic Models, Hidden Markov Process, Graphical Models, and Belief

Propagation 301
9.1 Topic Models ............................................................................................................... 301
9.2 Hidden Markov Model .............................................................................................. 305
9.3 Graphical Models, and Belief Propagation ......................................................... 310
9.4 Bayesian or Belief Networks .................................................................................... 311
9.5 Markov Random Fields .............................................................................................. 312
9.6 Factor Graphs ............................................................................................................... 313
9.7 Tree Algorithms ........................................................................................................... 314
9.8 Message Passing in general Graphs ....................................................................... 315
9.9 Graphs with a Single Cycle .................................................................................... 317

4
 9.10 Belief Update in Networks with aSingle Loop .................................................... 319
9.11 Maximum Weight Matching ................................................................................... 320
9.12 Warning Propagation ................................................................................................ 324
9.13 Correlation Between Variables ................................................................................ 325
9.14 Exercises ........................................................................................................................ 330

10 Other Topics 332

10.1 Rankings ........................................................................................................................ 332
10.2 Hare System for Voting ............................................................................................. 334
10.3 Compressed Sensing and SparseVectors ............................................................. 335
10.3.1 Unique Reconstruction of a Sparse Vector ............................................ 336
10.3.2 The Exact Reconstruction Property ......................................................... 339
10.3.3 Restricted Isometry Property ................................................................... 340
10.4 Applications ................................................................................................................. 342
10.4.1 Sparse Vector in Some Coordinate Basis ............................................... 342
10.4.2 A Representation Cannot be Sparse in Both Time and Frequency
Domains .............................................................................................. 342
10.4.3 Biological ....................................................................................................... 345
10.4.4 Finding Overlapping Cliques or Communities ................................... 345
10.4.5 Low Rank Matrices ...................................................................................... 346
10.5 Gradient ........................................................................................................................ 347
10.6 Linear Programming ................................................................................................ 348
10.6.1 The Ellipsoid Algorithm ............................................................................. 350
10.7 Integer Optimization ................................................................................................ 351
10.8 Semi-Denite Programming ................................................................................... 352
10.9 Exercises ........................................................................................................................ 354

11 Appendix 357
11.1 Asymptotic Notation ................................................................................................ 357
11.2 Useful relations .......................................................................................................... 358
11.3 Useful Inequalities .................................................................................................... 362
11.4 Probability ................................................................................................................. 369
11.4.1 Sample Space, Events, Independence .................................................... 370
11.4.2 Linearity of Expectation ............................................................................ 371
11.4.3 Union Bound ................................................................................................ 371
11.4.4 Indicator Variables ...................................................................................... 371
11.4.5 Variance .......................................................................................................... 372
11.4.6 Variance of the Sum of Independent Random Variables .................... 372
11.4.7 Median .......................................................................................................... 373
11.4.8 The Central Limit Theorem ..................................................................... 373
11.4.9 Probability Distributions ............................................................................ 373
11.4.10 Bayes Rule and Estimators .......................................................................... 376
11.4.11 Tail Bounds and Cherno inequalities ...................................................... 378

5
 11.5 Eigenvalues and Eigenvectors ............................................................................... 382
11.5.1 Eigenvalues and Eigenvectors 382
11.5.2 Symmetric Matrices 384
11.5.3 Relationship between SVD and Eigen Decomposition ....................... 386
11.5.4 Extremal Properties of Eigenvalues ......................................................... 386
11.5.5 Eigenvalues of the Sum of Two Symmetric Matrices ........................... 388
11.5.6 Norms .............................................................................................................. 390
11.5.7 Important Norms and Their Properties .................................................. 391
11.5.8 Linear Algebra ............................................................................................. 393
11.5.9 Distance between subspaces ...................................................................... 395
11.6 Generating Functions ............................................................................................. 396
11.6.1 Generating Functions for Sequences Dened by Recurrence Rela­
tionships .............................................................................................. 397
11.6.2 The Exponential Generating Function and the Moment Generating
Function .............................................................................................. 399
11.7 Miscellaneous .............................................................................................................. 401
11.7.1 Lagrange multipliers ................................................................................... 401
11.7.2 Finite Fields .................................................................................................... 401
11.7.3 Hash Functions ............................................................................................. 402
11.7.4 Application of Mean Value Theorem ..................................................... 402
11.7.5 Sperner's Lemma .......................................................................................... 403
11.7.6 Prufer................................................................................................................ 404
11.8 Exercises ........................................................................................................................ 405

Index 410

6
 Foundations of Data Sciencey

John Hopcroft and Ravindran Kannan

21/8/2014

1 Introduction
Computer science as an academic discipline began in the 60's. Emphasis was on pro­
gramming languages, compilers, operating systems, and the mathematical theory that
supported these areas. Courses in theoretical computer science covered nite automata,
regular expressions, context free languages, and computability. In the 70's, algorithms
was added as an important component of theory. The emphasis was on making computers
useful. Today, a fundamental change is taking place and the focus is more on applications.
There are many reasons for this change. The merging of computing and communications
has played an important role. The enhanced ability to observe, collect and store data in
the natural sciences, in commerce, and in other elds calls for a change in our understand­
ing of data and how to handle it in the modern setting. The emergence of the web and
social networks, which are by far the largest such structures, presents both opportunities
and challenges for theory.

While traditional areas of computer science are still important and highly skilled indi­
viduals are needed in these areas, the majority of researchers will be involved with using
computers to understand and make usable massive data arising in applications, not just
how to make computers useful on specic well-dened problems. With this in mind we
have written this book to cover the theory likely to be useful in the next 40 years, just as
automata theory, algorithms and related topics gave students an advantage in the last 40
years. One of the major changes is the switch from discrete mathematics to more of an
emphasis on probability, statistics, and numerical methods.

Early drafts of the book have been used for both undergraduate and graduate courses.
Background material needed for an undergraduate course has been put in the appendix.
For this reason, the appendix has homework problems.

This book starts with the treatment of high dimensional geometry. Modern data in
diverse elds such as Information Processing, Search, Machine Learning, etc., is often
yCopyright 2011. All rights reserved

7
represented advantageously as vectors with a large number of components. This is so
even in cases when the vector representation is not the natural rst choice. Our intuition
from two or three dimensional space can be surprisingly o the mark when it comes to
high dimensional space. Chapter 2 works out the fundamentals needed to understand the
dierences. The emphasis of the chapter, as well as the book in general, is to get across
the mathematical foundations rather than dwell on particular applications that are only
briey described.

The mathematical areas most relevant to dealing with high-dimensional data are ma­
trix algebra and algorithms. We focus on singular value decomposition, a central tool in
this area. Chapter 4 gives a from-rst-principles description of this. Applications of sin­
gular value decomposition include principal component analysis, a widely used technique
which we touch upon, as well as modern applications to statistical mixtures of probability
densities, discrete optimization, etc., which are described in more detail.

Central to our understanding of large structures, like the web and social networks, is
building models to capture essential properties of these structures. The simplest model is
that of a random graph formulated by Erdos and Renyi, which we study in detail proving
that certain global phenomena, like a giant connected component, arise in such structures
with only local choices. We also describe other models of random graphs.

One of the surprises of computer science over the last two decades is that some domain­
independent methods have been immensely successful in tackling problems from diverse
areas. Machine learning is a striking example. We describe the foundations of machine
learning, both learning from given training examples, as well as the theory of Vapnik-
Chervonenkis dimension, which tells us how many training examples suce for learning.
Another important domain-independent technique is based on Markov chains. The un­
derlying mathematical theory, as well as the connections to electrical networks, forms the
core of our chapter on Markov chains.

The eld of algorithms has traditionally assumed that the input data to a problem
is presented in random access memory, which the algorithm can repeatedly access. This
is not feasible for modern problems. The streaming model and other models have been
formulated to better reect this. In this setting, sampling plays a crucial role and, indeed,
we have to sample on the y. in Chapter ?? we study how to draw good samples eciently
and how to estimate statistical, as well as linear algebra quantities, with such samples.

One of the most important tools in the modern toolkit is clustering, dividing data into
groups of similar objects. After describing some of the basic methods for clustering, such
as the k-means algorithm, we focus on modern developments in understanding these, as
well as newer algorithms. The chapter ends with a study of clustering criteria.

This book also covers graphical models and belief propagation, ranking and voting,

8
sparse vectors, and compressed sensing. The appendix includes a wealth of background
material.

A word about notation in the book. To help the student, we have adopted certain
notations, and with a few exceptions, adhered to them. We use lower case letters for
scaler variables and functions, bold face lower case for vectors, and upper case letters
for matrices. Lower case near the beginning of the alphabet tend to be constants, in the
middle of the alphabet, such as i, j , and k; are indices in summations, n and m for integer
sizes, and x, y and z for variables. Where the literature traditionally uses a symbol for
a quantity, we also used that symbol, even if it meant abandoning our convention. If we
have a set of points in some vector space, and work with a subspace, we use n for the
number of points, d for the dimension of the space, and k for the dimension of the subspace.

The term "almost surely" means with probability one. We use ln n for the natural
logarithm and log n for the base two logarithm. If we want base ten, we will use log10 :
To simplify notation and to make it easier to read we use E2(1 — x) for (E(1 — x) and
E(1 — x)2 for E ((1 — x)2) :

9
2 High-Dimensional Space
In many applications data is in the form of vectors. In other applications, data is not
in the form of vectors, but could be usefully represented by vectors. The Vector Space
Model [SWY75] is a good example. In the vector space model, a document is represented
by a vector, each component of which corresponds to the number of occurrences of a par­
ticular term in the document. The English language has on the order of 25,000 words or
terms, so each document is represented by a 25,000 dimensional vector. A collection of n
documents is represented by a collection of n vectors, one vector per document. The vec­
tors may be arranged as columns of a 25; 000 n matrix. See Figure 2.1. A query is also
represented by a vector in the same space. The component of the vector corresponding to
a term in the query, species the importance of the term to the query. To nd documents
about cars that are not race cars, a query vector will have a large positive component for
the word car and also for the words engine and perhaps door, and a negative component
for the words race, betting, etc.

One needs a measure of relevance or similarity of a query to a document. The dot

product or cosine of the angle between the two vectors is an often used measure of sim­
ilarity. To respond to a query, one computes the dot product or the cosine of the angle
between the query vector and each document vector and returns the documents with the
highest values of these quantities. While it is by no means clear that this approach will
do well for the information retrieval problem, many empirical studies have established the
eectiveness of this general approach.

The vector space model is useful in ranking or ordering a large collection of documents
in decreasing order of importance. For large collections, an approach based on human
understanding of each document is not feasible. Instead, an automated procedure is
needed that is able to rank documents with those central to the collection ranked highest.
Each document is represented as a vector with the vectors forming the columns of a matrix
A. The similarity of pairs of documents is dened by the dot product of the vectors. All
pairwise similarities are contained in the matrix product ATA. If one assumes that the
documents central to the collection are those with high similarity to other documents, then
computing AT A enables one to create a ranking. Dene the total similarity of document
i to be the sum of the entries in the ith row of ATA and rank documents by their total
similarity. It turns out that with the vector representation on hand, a better way of
ranking is to rst nd the best t direction. That is, the unit vector u; for which the sum
of squared perpendicular distances of all the vectors to u is minimized. See Figure 2.2.
Then, one ranks the vectors according to their dot product with u. The best-t direction
is a well-studied notion in linear algebra. There is elegant theory and ecient algorithms
presented in Chapter 3 that facilitate the ranking as well as applications in many other
domains.

In the vector space representation of data, properties of vectors such as dot products,

10
Figure 2.1: A document and its term-document vector along with a collection of docu­
ments represented by their term-document vectors.

distance between vectors, and orthogonality, often have natural interpretations and this
is what makes the vector representation more important than just a book keeping device.
For example, the squared distance between two 0-1 vectors representing links on web pages
is the number of web pages linked to by only one of the pages. In Figure 2.3, pages 4 and
5 both have links to pages 1, 3, and 6, but only page 5 has a link to page 2. Thus, the
squared distance between the two vectors is one. We have seen that dot products measure
similarity. Orthogonality of two nonnegative vectors says that they are disjoint. Thus, if
a document collection, e.g., all news articles of a particular year, contained documents on
two or more disparate topics, vectors corresponding to documents from dierent topics
would be nearly orthogonal.

The dot product, cosine of the angle, distance, etc., are all measures of similarity or
dissimilarity, but there are important mathematical and algorithmic dierences between
them. The random projection theorem presented in this chapter states that a collection
of vectors can be projected to a lower-dimensional space approximately preserving all
pairwise distances between vectors. Thus, the nearest neighbors of each vector in the
collection can be computed in the projected lower-dimensional space. Such a savings in
time is not possible for computing pairwise dot products using a simple projection.

Our aim in this book is to present the reader with the mathematical foundations to
deal with high-dimensional data. There are two important parts of this foundation. The
rst is high-dimensional geometry, along with vectors, matrices, and linear algebra. The
second more modern aspect is the combination with probability.

High dimensionality is a common characteristic in many models and for this reason
much of this chapter is devoted to the geometry of high-dimensional space, which is quite
dierent from our intuitive understanding of two and three dimensions. We focus rst
on volumes and surface areas of high-dimensional objects like hyperspheres. We will not
present details of any one application, but rather present the fundamental theory useful
to many applications.

One reason probability comes in is that many computational problems are hard if
our algorithms are required to be ecient on all possible data. In practical situations,
domain knowledge often enables the expert to formulate stochastic models of data. In

11
 best t line

Figure 2.2: The best t line is the line that minimizes the sum of the squared perpendicular
distances.

Figure 2.3: Two web pages as vectors. The squared distance between the two vectors is
the number of web pages linked to by just one of the two web pages.

customer-product data, a common assumption is that the goods each customer buys
are independent of what goods the others buy. One may also assume that the goods
a customer buys satises a known probability law, like the Gaussian distribution. In
keeping with the spirit of the book, we do not discuss specic stochastic models, but
present the fundamentals. An important fundamental is the law of large numbers that
states that under the assumption of independence of customers, the total consumption
of each good is remarkably close to its mean value. The central limit theorem is of a
similar avor. Indeed, it turns out that picking random points from geometric objects
like hyperspheres exhibits almost identical properties in high dimensions. One calls this
phenomena the \law of large dimensions". We will establish these geometric properties
rst before discussing Cherno bounds and related theorems on aggregates of independent
random variables.

2.1 Properties of High-Dimensional Space

Our intuition about space was formed in two and three dimensions and is often mis­
leading in high dimensions. Consider placing 100 points uniformly at random in a unit
square. Each coordinate is generated independently and uniformly at random from the
interval [0, 1]. Select a point and measure the distance to all other points and observe

12
the distribution of distances. Then increase the dimension and generate the points uni­
formly at random in a 100-dimensional unit cube. The distribution of distances becomes
concentrated about an average distance. The reason is easy to see. Let x and y be two
such points in d-dimensions. The distance between x and y is
ud
jx_ yj = t 52 (xi_ y*)2.
i=1

Since Pd=1 (xi — yi)2 is the summation of a number of independent random variables of
bounded variance, by the law of large numbers the distribution of jx — yj2 is concentrated
about its expected value. Contrast this with the situation where the dimension is two or
three and the distribution of distances is spread out.

For another example, consider the dierence between picking a point uniformly at
random from a unit-radius circle and from a unit-radius sphere in d-dimensions. In d-
dimensions the distance from the point to the center of the sphere is very likely to be
between 1 — dd and 1, where c is a constant independent of d. This implies that most of
the mass is near the surface of the sphere. Furthermore, the rst coordinate, x1; of such a
point is likely to be between — pcd and + pcd, which we express by saying that most of the
mass is near the equator. The equator perpendicular to the x1 axis is the set fxjx1 = 0g.
We will prove these results in this chapter, but rst a review of some probability.

2.2 The Law of Large Numbers

In the previous section, we claimed that points generated at random in high dimen­
sions were all essentially the same distance apart. The reason is that if one averages
n independent samples x1; x2; : : : ;xn of a random variable x; the result will be close to
the expected value of x: Specically the probability that the average will dier from the
expected value by more than e is less than some value .
2
xi + x2 +----- + x-
Prob - E(x) (2.1)
n
Here the 2 in the numerator is the variance of x. The larger the variance of the random
variable, the greater the probability that the error will exceed e. The number of points n
is in the denominator since the more values that are averaged, the smaller the probability
that the dierence will exceed e. Similarly the larger e is, the smaller the probability that
the dierence will exceed e and hence e is in the denominator. Notice that squaring e
makes the fraction a dimensionalless quantity.

To prove the law of large numbers we use two inequalities. The rst is Markov's
inequality. One can bound the probability that a nonnegative random variable exceeds a
by the expected value of the variable divided by a.

13
 Theorem 2.1 (Markov's inequality) Let x be a nonnegative random variable. Then
for a > 0,
E E E x E(x)
Prob(x > a) < ------ .
a
Proof: We prove the theorem for continuous random variables. So we use integrals. The
same proof works for discrete random variables with sums instead of integrals.

() = p(x)dx =xp(x)dx +xp(x)dx >xp(x)dx

0 0aa

>Za ap(x)dx =aZa p(x)dx =ap(x >a)

Thus, Prob(x > a) < Ex).

Corollary 2.2 Prob (x > cE(x)) < C

Proof: Substitute cE (x) for a.

Markov's inequality bounds the tail of a distribution using only information about the
mean. A tighter bound can be obtained by also using the variance.

Theorem 2.3 (Chebyshev's inequality) Let x be a random variable with mean m and
variance 2 . Then
Prob(jx — mj > aa) < —.
a2
Proof: Prob(jx — mj > aa) = Prob ((x — m)2 > a2a2). Note that (x — m)2 is a nonneg­
ative random variable, so Markov's inequality can be a applied giving:

Prob ((x - m)2 > 2 ■■ < E «x ~ m>2) = " = 1

Probl(x m) > ' < a2a2 a2 a2 a2:

Thus, Prob (jx — mj > aa) < -22.

The law of large numbers follows from Chebyshev's inequality. Recall that E(x+y) =
E(x) + E(y); a2(cx) = c2a2(x); a2(x — m) = a2(x); and if x and y are independent, then
E (xy) = E(x)E(y) and a2 (x + y) = a2 (x) + a2 (y). To prove a2 (x + y) = a2(x) + a2(y)
when x and y are independent, since a2(x — m) = a2(x); one can assume E(x) = 0 and
E(y) = 0. Thus,

a2(x + y) = E ((x + y)2) = E(x2) + E(y2) + 2E(xy)

= E(x2) + E(y2) + 2E(x)E(y) = a2(x) + a2(y).

Replacing E(xy) by E(x)E(y) required independence.

14
Figure 2.4: Illustration of the relationship between the sphere and the cube in 2, 4, and
d-dimensions.

Theorem 2.4 (Law of large numbers) Let x1; x2;:: : ;xn be n samples of a random
variable x: Then

X1 + X2 + ’ ’ ’ + Xn
- E (x)
n

Proof: By Chebychev's inequality

x1 + x2 + + xn ^2 X X1 +X2+---- +Xn A
\_______ n_______ J_
Prob — E(x) > e
n e2

2 2 a2(xi + X2 +------- + xn)

n2e2
—2^ (a2(xi) + a2(x2) +------ + a2(xn))
n2e2
a2(x)
ne2

The law of large numbers bounds the dierence of the sample average and the expected
value. Note that the size of the sample for a given error bound is independent of the size
of the population class. In the limit, when the sample size goes to innity, the central
limit theorem says that the distribution of the sample average is Gaussian provided the
random variable has nite variance. Later, we will consider random variables that are
the sum of random variables. That is, x = xi + x2 + + xn: Cherno bounds will tell
us about the probability of x diering from its expected value. We will delay this until
Section .4..

2.3 The High-Dimensional Sphere

One of the interesting facts about a unit-radius sphere in high dimensions is that
as the dimension increases, the volume of the sphere goes to zero. This has important

5
 Figure 2.5: Conceptual drawing of a sphere and a cube.

implications. Also, the volume of a high-dimensional sphere is essentially all contained in

a thin slice at the equator and simultaneously in a narrow annulus at the surface. There is
essentially no interior volume. Similarly, the surface area is essentially all at the equator.
These facts, which are contrary to our two or three-dimensional intuition, will be proved
by integration.

2.3.1 The Sphere and the Cube in High Dimensions

Consider the dierence between the volume of a cube with unit-length sides and the
volume of a unit-radius sphere as the dimension d of the space increases. As the dimen­
sion of the cube increases, its volume is always one and the maximum possible distance
between two points grows as Vd. In contrast, as the dimension of a unit-radius sphere
increases, its volume goes to zero and the maximum possible distance between two points
stays at two.

For d=2, the unit square centered at the origin lies completely inside the unit-radius
circle. The distance from the origin to a vertex of the square is
//i\2 /1 \2 p pH
V( 1 ) +( 1 ) = "22 = 0:707:

Here, the square lies inside the circle. At d=4, the distance from the origin to a vertex of
a unit cube centered at the origin is

/nT2 71T2 71T2 71T2

V( 1 ) +( 1 ) +( 2 ) +( 1 ) = 1:

Thus, the vertex lies on the surface of the unit 4-sphere centered at the origin. As the
dimension d increases, the distance from the origin to a vertex of the cube increases as
'2d; and for large d, the vertices of the cube lie far outside the unit radius sphere. Figure

2.5 illustrates conceptually a cube and a sphere. The vertices of the cube are at distance

16
 Figure 2.6 : Volume of sphere in 2 and 3 dimensions.

from the origin and for large d lie outside the unit sphere. On the other hand, the mid
point of each face of the cube is only distance 1/2 from the origin and thus is inside the
sphere. For large d, almost all the volume of the cube is located outside the sphere.

2.3.2 Volume and Surface Area of the Unit Sphere

For xed dimension d, the volume of a sphere is a function of its radius and grows as
rd . For xed radius, the volume of a sphere is a function of the dimension of the space.
What is interesting is that the volume of a unit sphere goes to zero as the dimension of
the sphere increases.

To calculate the volume of a unit-radius sphere, one can integrate in either Cartesian
or polar coordinates. In Cartesian coordinates the volume of a unit sphere is given by

xd=V1-X2---- xd-1

■ I______ dxd •••dx2dx1.

xd=~y/ 1-x1------- x2

Since the limits of the integrals are complicated, it is easier to integrate using polar
coordinates. First, lets work out what happens in polar coordinates for d = 2 and d = 3.
[See Figure (2.6).] If d = 2, the volume is really the area (which we know to be ).
Consider a innitesimal radial triangle with the origin as the apex. The area between r
and r + dr of this triangle is bounded by two parallel arcs and two radial lines and since
the (innitesimal) arcs are perpendicular to the radius, the area of this piece is just ddr,
where, d is the arc length. In three dimensions, d is the area (2-dimensional volume)
and again, the surface of d is perpendicular to the radial direction, so the volume of the
piece is ddr.
In polar coordinates, V (d) is given by
1
V(d)=ZZ
rd 1drdQ.

Sd r=0

Here, d is the surface area of the innitesimal piece of the solid angle Sd of the unit
sphere. See Figure 2.7. The convex hull of the d piece and the origin form a cone. At
radius r, the surface area of the top of the cone is rd-1dQ since the surface area is d — 1
dimensional and each dimension scales by r. The volume of the innitesimal piece is base
times height, and since the surface of the sphere is perpendicular to the radial direction
at each point, the height is dr giving the above integral.

17
 Figure 2.7 : Innitesimal volume in a d-dimensional sphere of unit radius.

Since the variables and r do not interact,

1
r rd-1 A(d)
d =
d
Sd r=0 Sd

where A(d) is the surface area of a d-dimensional unit-radius sphere. The question re­
mains, how to determine the surface area A (d) = d:
Sd

Consider a dierent integral

xd dx2dx1:

Including the exponential allows integration to innity rather than stopping at the surface
of the sphere. Thus, I(d) can be computed by integrating in both Cartesian and polar
coordinates. Integrating in polar coordinates will relate I(d) to the surface area A(d).
Equating the two results for I(d) allows one to solve for A(d).

First, calculate I(d) by integration in Cartesian coordinates.

( rd d
I(d)= = ( = ^2 :

Here, we have used the fact that R1 e-x2 dx = p%. For a proof of this, see Section ??
of the appendix. Next, calculate I(d) by integrating in polar coordinates. The volume of
the differential element is rd-1dQdr. Thus,

e r2 rd 1dr.

18
Cartesian coordinates

I(d)=

too hard because of limits

evaluate I(d) instead

Polar coordinates m equate and solve for A(d)

1
A(d) I (d) = y dtif e~rr rd~1dr = A(d)1r
rd 1dr
d
Sd 0

substitute value of A(d)

into formula for V (d)

Equate integrals for I(d) in Cartesian and polar coordinates and solve for A(d):
Substitute A(d) into the formula for volume of the sphere obtained by integrating in
polar coordinates. This gives the result for V (d):

Figure 2.8 : Strategy for calculating the volume of a d-dimensional sphere.

The integral d is the integral over the entire solid angle and gives the surface area,
Sd

A(d), of a unit sphere. Thus, I (d) = A (d) f e~r2rd~1dr. Evaluating the remaining
integral gives
7 e-r2rd-1dr =1 7 e"ttd “ 1 dt = 1r (d^
2 2J 2 \2)
00

and hence, I(d) = A(d)1r (d) where the gamma function r (x) is a generalization of the
factorial function for noninteger values of x. r (x) = (x — 1) r (x — 1), T (1) = T (2) = 1,
and r Q) v'^. For integer x, T (x) = (x — 1)!.

Combining I (d) = k 2 with I (d) = A (d) 11 r (d) yields

d
k2
A (d)

establishing the following lemma.

Lemma 2.5 The surface area A(d) and the volume V (d) of a unit-radius sphere in d

19
dimensions are given by

2 K2
A (d) and V (d) = :
d r (d)

To check the formula for the volume of a unit sphere, note that V (2) = and
3
V (3) = fzpk = 3k, which are the correct volumes for the unit spheres in two and
r( 2)
three dimensions. To check the formula for the surface area of a unit sphere, note that
3
A(2) = 2k and A(3) = IpZ = 4k, which are the correct surface areas for the unit sphere
in two and three dimensions. Note that kd is an exponential in d and r (2) grows as the
!1 V(d) = 0, as claimed.
factorial of 2d. This implies that dlim

The volume of a d-dimensional sphere of radius r grows as rd: This follows since the
unit sphere can be mapped to a sphere of radius r by the linear transformation specied
by a diagonal matrix with diagonal elements r: The determinant of this matrix is rd: See
Section 2.4. Since the surface area is the derivative of the volume, the surface area grows
as rd-1: See last paragraph of Section 2.3.5.
The proof of Lemma 2.5 illustrates the relationship between the surface area of the
sphere and the Gaussian probability density

1 e~(xi+x2+-----+xd)2=2
P2K

This relationship is an important one and will be used several times in this chapter.

2.3.3 The Volume is Near the Equator

Consider a high-dimensional unit-radius sphere and x the North Pole on the x1 axis
at x1 = 1. Divide the sphere in half by intersecting it with the plane x1 = 0. The
intersection of the plane with the sphere forms a region of one lower dimension, namely
{x jxj < 1;X1 = 0} ; called the equator. The intersection is a sphere of dimension d — 1
and has volume V (d — 1). In three dimensions this region is a circle, in four dimensions
the region is a 3-dimensional sphere, etc. In our terminology, a circle is a 2-dimensional
sphere and its volume is what one usually refers to as the area of a circle. The surface
area of the 2-dimensional sphere is what one usually refers to as the circumference of a
circle.

It turns out that essentially all of the volume of the upper hemisphere lies between
the plane x1 = 0 and a parallel plane, x1 = ", that is slightly higher. For what value of
" does essentially all the volume lie between x1 = 0 and x1 = "? The answer depends
on the dimension. For dimension d; it is O(pd^)• Before we prove this, some intuition
is in order. Since jxj2 = x21 + x2 + + x2d and by symmetry, we expect the xi2 's to be
generally equal (or close to each other), we expect each xi2 to be at most O(1=d). Now for

20
Discovering Diverse Content Through
Random Scribd Documents
 “Look at those flies on the ceiling. Why have you not drowned
them?”

“Huzoor, they return.”

“Like all evil things.”

To divert the conversation, Hassan related how the kitchen-boy

had killed a snake, good, but killed it by cutting it in two, bad,
because it becomes two snakes.

“When he breaks a plate, does it become two plates?”

“Glasses and a new teapot will similarly be required, also for

myself a coat.”

Aziz sighed. Each for himself. One man needs a coat, another a
rich wife; each approaches his goal by a clever detour. Fielding had
saved the girl a fine of twenty thousand rupees, and now followed
her to England. If he desired to marry her, all was explained; she
would bring him a larger dowry. Aziz did not believe his own
suspicions—better if he had, for then he would have denounced and
cleared the situation up. Suspicion and belief could in his mind exist
side by side. They sprang from different sources, and need never
intermingle. Suspicion in the Oriental is a sort of malignant tumour, a
mental malady, that makes him self-conscious and unfriendly
suddenly; he trusts and mistrusts at the same time in a way the
Westerner cannot comprehend. It is his demon, as the Westerner’s is
hypocrisy. Aziz was seized by it, and his fancy built a satanic castle,
of which the foundation had been laid when he talked at Dilkusha
under the stars. The girl had surely been Cyril’s mistress when she
stopped in the College—Mohammed Latif was right. But was that all?
Perhaps it was Cyril who followed her into the cave. . . . No;
impossible. Cyril hadn’t been on the Kawa Dol at all. Impossible.
Ridiculous. Yet the fancy left him trembling with misery. Such
treachery—if true—would have been the worst in Indian history;
nothing so vile, not even the murder of Afzul Khan by Sivaji. He was
shaken, as though by a truth, and told Hassan to leave him.

Next day he decided to take his children back to Mussoorie. They

had come down for the trial, that he might bid them farewell, and
had stayed on at Hamidullah’s for the rejoicings. Major Roberts
would give him leave, and during his absence Fielding would go off
to England. The idea suited both his beliefs and his suspicions.
Events would prove which was right, and preserve, in either case,
his dignity.

Fielding was conscious of something hostile, and because he was

really fond of Aziz his optimism failed him. Travelling light is less
easy as soon as affection is involved. Unable to jog forward in the
serene hope that all would come right, he wrote an elaborate letter
in the rather modern style: “It is on my mind that you think me a
prude about women. I had rather you thought anything else of me.
If I live impeccably now, it is only because I am well on the forties—
a period of revision. In the eighties I shall revise again. And before
the nineties come—I shall be revised! But, alive or dead, I am
absolutely devoid of morals. Do kindly grasp this about me.” Aziz did
not care for the letter at all. It hurt his delicacy. He liked
confidences, however gross, but generalizations and comparisons
always repelled him. Life is not a scientific manual. He replied coldly,
regretting his inability to return from Mussoorie before his friend
sailed: “But I must take my poor little holiday while I can. All must
be economy henceforward, all hopes of Kashmir have vanished for
ever and ever. When you return I shall be slaving far away in some
new post.”

And Fielding went, and in the last gutterings of Chandrapore—

heaven and earth both looking like toffee—the Indian’s bad fancies
were confirmed. His friends encouraged them, for though they had
liked the Principal, they felt uneasy at his getting to know so much
about their private affairs. Mahmoud Ali soon declared that treachery
was afoot. Hamidullah murmured, “Certainly of late he no longer
addressed us with his former frankness,” and warned Aziz “not to
expect too much—he and she are, after all, both members of
another race.” “Where are my twenty thousand rupees?” he thought.
He was absolutely indifferent to money—not merely generous with
it, but promptly paying his debts when he could remember to do so
—yet these rupees haunted his mind, because he had been tricked
about them, and allowed them to escape overseas, like so much of
the wealth of India. Cyril would marry Miss Quested—he grew
certain of it, all the unexplained residue of the Marabar contributing.
It was the natural conclusion of the horrible senseless picnic, and
before long he persuaded himself that the wedding had actually
taken place.
 CHAPTER XXXII
Egypt was charming—a green strip of carpet and walking up and
down it four sorts of animals and one sort of man. Fielding’s
business took him there for a few days. He re-embarked at
Alexandria—bright blue sky, constant wind, clean low coast-line, as
against the intricacies of Bombay. Crete welcomed him next with the
long snowy ridge of its mountains, and then came Venice. As he
landed on the piazzetta a cup of beauty was lifted to his lips, and he
drank with a sense of disloyalty. The buildings of Venice, like the
mountains of Crete and the fields of Egypt, stood in the right place,
whereas in poor India everything was placed wrong. He had
forgotten the beauty of form among idol temples and lumpy hills;
indeed, without form, how can there be beauty? Form stammered
here and there in a mosque, became rigid through nervousness
even, but oh these Italian churches! San Giorgio standing on the
island which could scarcely have risen from the waves without it, the
Salute holding the entrance of a canal which, but for it, would not be
the Grand Canal! In the old undergraduate days he had wrapped
himself up in the many-coloured blanket of St. Mark’s, but something
more precious than mosaics and marbles was offered to him now:
the harmony between the works of man and the earth that upholds
them, the civilization that has escaped muddle, the spirit in a
reasonable form, with flesh and blood subsisting. Writing picture
post-cards to his Indian friends, he felt that all of them would miss
the joys he experienced now, the joys of form, and that this
constituted a serious barrier. They would see the sumptuousness of
Venice, not its shape, and though Venice was not Europe, it was part
of the Mediterranean harmony. The Mediterranean is the human
norm. When men leave that exquisite lake, whether through the
Bosphorus or the Pillars of Hercules, they approach the monstrous
and extraordinary; and the southern exit leads to the strangest
experience of all. Turning his back on it yet again, he took the train
northward, and tender romantic fancies that he thought were dead
for ever, flowered when he saw the buttercups and daisies of June.
PART III: TEMPLE
 CHAPTER XXXIII
Some hundreds of miles westward of the Marabar Hills, and two
years later in time, Professor Narayan Godbole stands in the
presence of God. God is not born yet—that will occur at midnight—
but He has also been born centuries ago, nor can He ever be born,
because He is the Lord of the Universe, who transcends human
processes. He is, was not, is not, was. He and Professor Godbole
stood at opposite ends of the same strip of carpet.

“Tukaram, Tukaram,
Thou art my father and mother and everybody.
Tukaram, Tukaram,
Thou art my father and mother and everybody.
Tukaram, Tukaram,
Thou art my father and mother and everybody.
Tukaram, Tukaram,
Thou art my father and mother and everybody.
Tukaram. . . .”

This corridor in the palace at Mau opened through other corridors

into a courtyard. It was of beautiful hard white stucco, but its pillars
and vaulting could scarcely be seen behind coloured rags, iridescent
balls, chandeliers of opaque pink glass, and murky photographs
framed crookedly. At the end was the small but famous shrine of the
dynastic cult, and the God to be born was largely a silver image the
size of a teaspoon. Hindus sat on either side of the carpet where
they could find room, or overflowed into the adjoining corridors and
the courtyard—Hindus, Hindus only, mild-featured men, mostly
villagers, for whom anything outside their villages passed in a
dream. They were the toiling ryot, whom some call the real India.
Mixed with them sat a few tradesmen out of the little town, officials,
courtiers, scions of the ruling house. Schoolboys kept inefficient
order. The assembly was in a tender, happy state unknown to an
English crowd, it seethed like a beneficent potion. When the villagers
broke cordon for a glimpse of the silver image, a most beautiful and
radiant expression came into their faces, a beauty in which there
was nothing personal, for it caused them all to resemble one another
during the moment of its indwelling, and only when it was
withdrawn did they revert to individual clods. And so with the music.
Music there was, but from so many sources that the sum-total was
untrammelled. The braying banging crooning melted into a single
mass which trailed round the palace before joining the thunder. Rain
fell at intervals throughout the night.

It was the turn of Professor Godbole’s choir. As Minister of

Education, he gained this special honour. When the previous group
of singers dispersed into the crowd, he pressed forward from the
back, already in full voice, that the chain of sacred sounds might be
uninterrupted. He was barefoot and in white, he wore a pale blue
turban; his gold pince-nez had caught in a jasmine garland, and lay
sideways down his nose. He and the six colleagues who supported
him clashed their cymbals, hit small drums, droned upon a portable
harmonium, and sang:

“Tukaram, Tukaram,
Thou art my father and mother and everybody.
Tukaram, Tukaram,
Thou art my father and mother and everybody.
Tukaram, Tukaram. . . .”

They sang not even to the God who confronted them, but to a
saint; they did not one thing which the non-Hindu would feel
dramatically correct; this approaching triumph of India was a muddle
(as we call it), a frustration of reason and form. Where was the God
Himself, in whose honour the congregation had gathered?
Indistinguishable in the jumble of His own altar, huddled out of sight
amid images of inferior descent, smothered under rose-leaves,
overhung by oleographs, outblazed by golden tablets representing
the Rajah’s ancestors, and entirely obscured, when the wind blew,
by the tattered foliage of a banana. Hundreds of electric lights had
been lit in His honour (worked by an engine whose thumps
destroyed the rhythm of the hymn). Yet His face could not be seen.
Hundreds of His silver dishes were piled around Him with the
minimum of effect. The inscriptions which the poets of the State had
composed were hung where they could not be read, or had twitched
their drawing-pins out of the stucco, and one of them (composed in
English to indicate His universality) consisted, by an unfortunate slip
of the draughtsman, of the words, “God si Love.”

God si Love. Is this the first message of India?

“Tukaram, Tukaram . . .,”

continued the choir, reinforced by a squabble behind the purdah

curtain, where two mothers tried to push their children at the same
moment to the front. A little girl’s leg shot out like an eel. In the
courtyard, drenched by the rain, the small Europeanized band
stumbled off into a waltz. “Nights of Gladness” they were playing.
The singers were not perturbed by this rival, they lived beyond
competition. It was long before the tiny fragment of Professor
Godbole that attended to outside things decided that his pince-nez
was in trouble, and that until it was adjusted he could not choose a
new hymn. He laid down one cymbal, with the other he clashed the
air, with his free hand he fumbled at the flowers round his neck. A
colleague assisted him. Singing into one another’s grey moustaches,
they disentangled the chain from the tinsel into which it had sunk.
Godbole consulted the music-book, said a word to the drummer, who
broke rhythm, made a thick little blur of sound, and produced a new
rhythm. This was more exciting, the inner images it evoked more
definite, and the singers’ expressions became fatuous and languid.
They loved all men, the whole universe, and scraps of their past, tiny
splinters of detail, emerged for a moment to melt into the universal
warmth. Thus Godbole, though she was not important to him,
remembered an old woman he had met in Chandrapore days.
Chance brought her into his mind while it was in this heated state,
he did not select her, she happened to occur among the throng of
soliciting images, a tiny splinter, and he impelled her by his spiritual
force to that place where completeness can be found. Completeness,
not reconstruction. His senses grew thinner, he remembered a wasp
seen he forgot where, perhaps on a stone. He loved the wasp
equally, he impelled it likewise, he was imitating God. And the stone
where the wasp clung—could he . . . no, he could not, he had been
wrong to attempt the stone, logic and conscious effort had seduced,
he came back to the strip of red carpet and discovered that he was
dancing upon it. Up and down, a third of the way to the altar and
back again, clashing his cymbals, his little legs twinkling, his
companions dancing with him and each other. Noise, noise, the
Europeanized band louder, incense on the altar, sweat, the blaze of
lights, wind in the bananas, noise, thunder, eleven-fifty by his wrist-
watch, seen as he threw up his hands and detached the tiny
reverberation that was his soul. Louder shouts in the crowd. He
danced on. The boys and men who were squatting in the aisles were
lifted forcibly and dropped without changing their shapes into the
laps of their neighbours. Down the path thus cleared advanced a
litter. It was the aged ruler of the state, brought against the advice
of his physicians to witness the Birth ceremony.

No one greeted the Rajah, nor did he wish it; this was no moment
for human glory. Nor could the litter be set down, lest it defiled the
temple by becoming a throne. He was lifted out of it while its feet
remained in air, and deposited on the carpet close to the altar, his
immense beard was straightened, his legs tucked under him, a paper
containing red powder was placed in his hand. There he sat, leaning
against a pillar, exhausted with illness, his eyes magnified by many
unshed tears.

He had not to wait long. In a land where all else was unpunctual,
the hour of the Birth was chronometrically observed. Three minutes
before it was due, a Brahman brought forth a model of the village of
Gokul (the Bethlehem in that nebulous story) and placed it in front
of the altar. The model was on a wooden tray about a yard square; it
was of clay, and was gaily blue and white with streamers and paint.
Here, upon a chair too small for him and with a head too large, sat
King Kansa, who is Herod, directing the murder of some Innocents,
and in a corner, similarly proportioned, stood the father and mother
of the Lord, warned to depart in a dream. The model was not holy,
but more than a decoration, for it diverted men from the actual
image of the God, and increased their sacred bewilderment. Some of
the villagers thought the Birth had occurred, saying with truth that
the Lord must have been born, or they could not see Him. But the
clock struck midnight, and simultaneously the rending note of the
conch broke forth, followed by the trumpeting of elephants; all who
had packets of powder threw them at the altar, and in the rosy dust
and incense, and clanging and shouts, Infinite Love took upon itself
the form of Shri Krishna, and saved the world. All sorrow was
annihilated, not only for Indians, but for foreigners, birds, caves,
railways, and the stars; all became joy, all laughter; there had never
been disease nor doubt, misunderstanding, cruelty, fear. Some
jumped in the air, others flung themselves prone and embraced the
bare feet of the universal lover; the women behind the purdah
slapped and shrieked; the little girl slipped out and danced by
herself, her black pigtails flying. Not an orgy of the body; the
tradition of that shrine forbade it. But the human spirit had tried by
a desperate contortion to ravish the unknown, flinging down science
and history in the struggle, yes, beauty herself. Did it succeed?
Books written afterwards say “Yes.” But how, if there is such an
event, can it be remembered afterwards? How can it be expressed in
anything but itself? Not only from the unbeliever are mysteries hid,
but the adept himself cannot retain them. He may think, if he
chooses, that he has been with God, but as soon as he thinks it, it
becomes history, and falls under the rules of time.

A cobra of papier-mâché now appeared on the carpet, also a

wooden cradle swinging from a frame. Professor Godbole
approached the latter with a red silk napkin in his arms. The napkin
was God, not that it was, and the image remained in the blur of the
altar. It was just a napkin, folded into a shape which indicated a
baby’s. The Professor dandled it and gave it to the Rajah, who,
making a great effort, said, “I name this child Shri Krishna,” and
tumbled it into the cradle. Tears poured from his eyes, because he
had seen the Lord’s salvation. He was too weak to exhibit the silk
baby to his people, his privilege in former years. His attendants lifted
him up, a new path was cleared through the crowd, and he was
carried away to a less sacred part of the palace. There, in a room
accessible to Western science by an outer staircase, his physician,
Dr. Aziz, awaited him. His Hindu physician, who had accompanied
him to the shrine, briefly reported his symptoms. As the ecstasy
receded, the invalid grew fretful. The bumping of the steam engine
that worked the dynamo disturbed him, and he asked for what
reason it had been introduced into his home. They replied that they
would enquire, and administered a sedative.

Down in the sacred corridors, joy had seethed to jollity. It was

their duty to play various games to amuse the newly born God, and
to simulate his sports with the wanton dairymaids of Brindaban.
Butter played a prominent part in these. When the cradle had been
removed, the principal nobles of the state gathered together for an
innocent frolic. They removed their turbans, and one put a lump of
butter on his forehead, and waited for it to slide down his nose into
his mouth. Before it could arrive, another stole up behind him,
snatched the melting morsel, and swallowed it himself. All laughed
exultantly at discovering that the divine sense of humour coincided
with their own. “God si love!” There is fun in heaven. God can play
practical jokes upon Himself, draw chairs away from beneath His
own posteriors, set His own turbans on fire, and steal His own
petticoats when He bathes. By sacrificing good taste, this worship
achieved what Christianity has shirked: the inclusion of merriment.
All spirit as well as all matter must participate in salvation, and if
practical jokes are banned, the circle is incomplete. Having
swallowed the butter, they played another game which chanced to
be graceful: the fondling of Shri Krishna under the similitude of a
child. A pretty red and gold ball is thrown, and he who catches it
chooses a child from the crowd, raises it in his arms, and carries it
round to be caressed. All stroke the darling creature for the Creator’s
sake, and murmur happy words. The child is restored to his parents,
the ball thrown on, and another child becomes for a moment the
World’s Desire. And the Lord bounds hither and thither through the
aisles, chance, and the sport of chance, irradiating little mortals with
His immortality. . . . When they had played this long enough—and
being exempt from boredom, they played it again and again, they
played it again and again—they took many sticks and hit them
together, whack smack, as though they fought the Pandava wars,
and threshed and churned with them, and later on they hung from
the roof of the temple, in a net, a great black earthenware jar, which
was painted here and there with red, and wreathed with dried figs.
Now came a rousing sport. Springing up, they struck at the jar with
their sticks. It cracked, broke, and a mass of greasy rice and milk
poured on to their faces. They ate and smeared one another’s
mouths, and dived between each other’s legs for what had been
pashed upon the carpet. This way and that spread the divine mess,
until the line of schoolboys, who had somewhat fended off the
crowd, broke for their share. The corridors, the courtyard, were filled
with benign confusion. Also the flies awoke and claimed their share
of God’s bounty. There was no quarrelling, owing to the nature of
the gift, for blessed is the man who confers it on another, he
imitates God. And those “imitations,” those “substitutions,” continued
to flicker through the assembly for many hours, awaking in each
man, according to his capacity, an emotion that he would not have
had otherwise. No definite image survived; at the Birth it was
questionable whether a silver doll or a mud village, or a silk napkin,
or an intangible spirit, or a pious resolution, had been born. Perhaps
all these things! Perhaps none! Perhaps all birth is an allegory! Still,
it was the main event of the religious year. It caused strange
thoughts. Covered with grease and dust, Professor Godbole had
once more developed the life of his spirit. He had, with increasing
vividness, again seen Mrs. Moore, and round her faintly clinging
forms of trouble. He was a Brahman, she Christian, but it made no
difference, it made no difference whether she was a trick of his
memory or a telepathic appeal. It was his duty, as it was his desire,
to place himself in the position of the God and to love her, and to
place himself in her position and to say to the God, “Come, come,
come, come.” This was all he could do. How inadequate! But each
according to his own capacities, and he knew that his own were
small. “One old Englishwoman and one little, little wasp,” he
thought, as he stepped out of the temple into the grey of a pouring
wet morning. “It does not seem much, still it is more than I am
myself.”
 CHAPTER XXXIV
Dr. Aziz left the palace at the same time. As he returned to his
house—which stood in a pleasant garden further up the main street
of the town—he could see his old patron paddling and capering in
the slush ahead. “Hullo!” he called, and it was the wrong remark, for
the devotee indicated by circular gestures of his arms that he did not
desire to be disturbed. He added, “Sorry,” which was right, for
Godbole twisted his head till it didn’t belong to his body, and said in
a strained voice that had no connection with his mind: “He arrived at
the European Guest House perhaps—at least possibly.”

“Did he? Since when?”

But time was too definite. He waved his arm more dimly and
disappeared. Aziz knew who “he” was—Fielding—but he refused to
think about him, because it disturbed his life, and he still trusted the
floods to prevent him from arriving. A fine little river issued from his
garden gate and gave him much hope. It was impossible that
anyone could get across from Deora in such weather as this.
Fielding’s visit was official. He had been transferred from
Chandrapore, and sent on a tour through Central India to see what
the remoter states were doing with regard to English education. He
had married, he had done the expected with Miss Quested, and Aziz
had no wish to see him again.

“Dear old Godbole,” he thought, and smiled. He had no religious

curiosity, and had never discovered the meaning of this annual antic,
but he was well assured that Godbole was a dear old man. He had
come to Mau through him and remained on his account. Without him
he could never have grasped problems so totally different from those
of Chandrapore. For here the cleavage was between Brahman and
non-Brahman; Moslems and English were quite out of the running,
and sometimes not mentioned for days. Since Godbole was a
Brahman, Aziz was one also for purposes of intrigue: they would
often joke about it together. The fissures in the Indian soil are
infinite: Hinduism, so solid from a distance, is riven into sects and
clans, which radiate and join, and change their names according to
the aspect from which they are approached. Study it for years with
the best teachers, and when you raise your head, nothing they have
told you quite fits. Aziz, the day of his inauguration, had remarked:
“I study nothing, I respect”—making an excellent impression. There
was now a minimum of prejudice against him. Nominally under a
Hindu doctor, he was really chief medicine man to the court. He had
to drop inoculation and such Western whims, but even at
Chandrapore his profession had been a game, centring round the
operating table, and here in the backwoods he let his instruments
rust, ran his little hospital at half steam, and caused no undue alarm.

His impulse to escape from the English was sound. They had
frightened him permanently, and there are only two reactions
against fright: to kick and scream on committees, or to retreat to a
remote jungle, where the sahib seldom comes. His old lawyer friends
wanted him to stop in British India and help agitate, and might have
prevailed, but for the treachery of Fielding. The news had not
surprised him in the least. A rift had opened between them after the
trial when Cyril had not joined in his procession; those advocacies of
the girl had increased it; then came the post-cards from Venice, so
cold, so unfriendly that all agreed that something was wrong; and
finally, after a silence, the expected letter from Hampstead.
Mahmoud Ali was with him at the time. “Some news that will
surprise you. I am to marry someone whom you know. . .” He did
not read further. “Here it comes, answer for me——” and he threw it
to Mahmoud Ali. Subsequent letters he destroyed unopened. It was
the end of a foolish experiment. And though sometimes at the back
of his mind he felt that Fielding had made sacrifices for him, it was
now all confused with his genuine hatred of the English. “I am an
Indian at last,” he thought, standing motionless in the rain.

Life passed pleasantly, the climate was healthy so that the children
could be with him all the year round, and he had married again—not
exactly a marriage, but he liked to regard it as one—and he read his
Persian, wrote his poetry, had his horse, and sometimes got some
shikar while the good Hindus looked the other way. His poems were
all on one topic—Oriental womanhood. “The purdah must go,” was
their burden, “otherwise we shall never be free.” And he declared
(fantastically) that India would not have been conquered if women
as well as men had fought at Plassy. “But we do not show our
women to the foreigner”—not explaining how this was to be
managed, for he was writing a poem. Bulbuls and roses would still
persist, the pathos of defeated Islam remained in his blood and
could not be expelled by modernities. Illogical poems—like their
writer. Yet they struck a true note: there cannot be a mother-land
without new homes. In one poem—the only one funny old Godbole
liked—he had skipped over the mother-land (whom he did not truly
love) and gone straight to internationality. “Ah, that is bhakti; ah, my
young friend, that is different and very good. Ah, India, who seems
not to move, will go straight there while the other nations waste
their time. May I translate this particular one into Hindi? In fact, it
might be rendered into Sanskrit almost, it is so enlightened. Yes, of
course, all your other poems are very good too. His Highness was
saying to Colonel Maggs last time he came that we are proud of
you”—simpering slightly.

Colonel Maggs was the Political Agent for the neighbourhood and
Aziz’ dejected opponent. The Criminal Investigation Department kept
an eye on Aziz ever since the trial—they had nothing actionable
against him, but Indians who have been unfortunate must be
watched, and to the end of his life he remained under observation,
thanks to Miss Quested’s mistake. Colonel Maggs learnt with concern
that a suspect was coming to Mau, and, adopting a playful manner,
rallied the old Rajah for permitting a Moslem doctor to approach his
sacred person. A few years ago, the Rajah would have taken the
hint, for the Political Agent then had been a formidable figure,
descending with all the thunders of Empire when it was most
inconvenient, turning the polity inside out, requiring motor-cars and
tiger-hunts, trees cut down that impeded the view from the Guest
House, cows milked in his presence, and generally arrogating the
control of internal affairs. But there had been a change of policy in
high quarters. Local thunders were no longer endorsed, and the
group of little states that composed the agency discovered this and
began comparing notes with fruitful result. To see how much, or how
little, Colonel Maggs would stand, became an agreeable game at
Mau, which was played by all the departments of State. He had to
stand the appointment of Dr. Aziz. The Rajah did not take the hint,
but replied that Hindus were less exclusive than formerly, thanks to
the enlightened commands of the Viceroy, and he felt it his duty to
move with the times.

Yes, all had gone well hitherto, but now, when the rest of the
state was plunged in its festival, he had a crisis of a very different
sort. A note awaited him at his house. There was no doubt that
Fielding had arrived overnight, nor much doubt that Godbole knew
of his arrival, for the note was addressed to him, and he had read it
before sending it on to Aziz, and had written in the margin, “Is not
this delightful news, but unfortunately my religious duties prevent
me from taking any action.” Fielding announced that he had
inspected Mudkul (Miss Derek’s former preserve), that he had nearly
been drowned at Deora, that he had reached Mau according to time-
table, and hoped to remain there two days, studying the various
educational innovations of his old friend. Nor had he come alone. His
wife and her brother accompanied him. And then the note turned
into the sort of note that always did arrive from the State Guest
House. Wanting something. No eggs. Mosquito nets torn. When
would they pay their respects to His Highness? Was it correct that a
torchlight procession would take place? If so, might they view it?
They didn’t want to give trouble, but if they might stand in a
balcony, or if they might go out in a boat. . . . Aziz tore the note up.
He had had enough of showing Miss Quested native life. Treacherous
hideous harridan! Bad people altogether. He hoped to avoid them,
though this might be difficult, for they would certainly be held up for
several days at Mau. Down country, the floods were even worse, and
the pale grey faces of lakes had appeared in the direction of the
Asirgarh railway station.
 CHAPTER XXXV
Long before he discovered Mau, another young Mohammedan had
retired there—a saint. His mother said to him, “Free prisoners.” So
he took a sword and went up to the fort. He unlocked a door, and
the prisoners streamed out and resumed their previous occupations,
but the police were too much annoyed and cut off the young man’s
head. Ignoring its absence, he made his way over the rocks that
separate the fort and the town, killing policemen as he went, and he
fell outside his mother’s house, having accomplished her orders.
Consequently there are two shrines to him to-day—that of the Head
above, and that of the Body below—and they are worshipped by the
few Mohammedans who live near, and by Hindus also. “There is no
God but God”; that symmetrical injunction melts in the mild airs of
Mau; it belongs to pilgrimages and universities, not to feudalism and
agriculture. When Aziz arrived, and found that even Islam was
idolatrous, he grew scornful, and longed to purify the place, like
Alamgir. But soon he didn’t mind, like Akbar. After all, this saint had
freed prisoners, and he himself had lain in prison. The Shrine of the
Body lay in his own garden and produced a weekly crop of lamps
and flowers, and when he saw them he recalled his sufferings. The
Shrine of the Head made a nice short walk for the children. He was
off duty the morning after the great pujah, and he told them to
come. Jemila held his hand. Ahmed and Karim ran in front, arguing
what the body looked like as it came staggering down, and whether
they would have been frightened if they met it. He didn’t want them
to grow up superstitious, so he rebuked them, and they answered
yes father, for they were well brought up, but, like himself, they were
impervious to argument, and after a polite pause they continued
saying what their natures compelled them to say.
 A slim, tall eight-sided building stood at the top of the slope,
among some bushes. This was the Shrine of the Head. It had not
been roofed, and was indeed merely a screen. Inside it crouched a
humble dome, and inside that, visible through a grille, was a
truncated gravestone, swathed in calico. The inner angles of the
screen were cumbered with bees’ nests, and a gentle shower of
broken wings and other aerial oddments kept falling, and had strewn
the damp pavement with their flue. Ahmed, apprized by Mohammed
Latif of the character of the bee, said, “They will not hurt us, whose
lives are chaste,” and pushed boldly in; his sister was more cautious.
From the shrine they went to a mosque, which, in size and design,
resembled a fire-screen; the arcades of Chandrapore had shrunk to
a flat piece of ornamental stucco, with protuberances at either end
to suggest minarets. The funny little thing didn’t even stand straight,
for the rock on which it had been put was slipping down the hill. It,
and the shrine, were a strange outcome of the protests of Arabia.

They wandered over the old fort, now deserted, and admired the
various views. The scenery, according to their standards, was
delightful—the sky grey and black, bellyfuls of rain all over it, the
earth pocked with pools of water and slimy with mud. A magnificent
monsoon—the best for three years, the tanks already full, bumper
crops possible. Out towards the river (the route by which the
Fieldings had escaped from Deora) the downpour had been
enormous, the mails had to be pulled across by ropes. They could
just see the break in the forest trees where the gorge came through,
and the rocks above that marked the site of the diamond mine,
glistening with wet. Close beneath was the suburban residence of
the Junior Rani, isolated by floods, and Her Highness, lax about
purdah, to be seen paddling with her handmaidens in the garden
and waving her sari at the monkeys on the roof. But better not look
close beneath, perhaps—nor towards the European Guest House
either. Beyond the Guest House rose another grey-green gloom of
hills, covered with temples like little white flames. There were over
two hundred gods in that direction alone, who visited each other
constantly, and owned numerous cows, and all the betel-leaf
industry, besides having shares in the Asirgarh motor omnibus. Many
of them were in the palace at this moment, having the time of their
lives; others, too large or proud to travel, had sent symbols to
represent them. The air was thick with religion and rain.

Their white shirts fluttering, Ahmed and Karim ran about over the
fort, shrieking with joy. Presently they intersected a line of prisoners,
who were looking aimlessly at an old bronze gun. “Which of you is to
be pardoned?” they asked. For to-night was the procession of the
Chief God, when He would leave the palace, escorted by the whole
power of the State, and pass by the Jail, which stood down in the
town now. As He did so, troubling the waters of our civilization, one
prisoner would be released, and then He would proceed to the great
Mau tank that stretched as far as the Guest House garden, where
something else would happen, some final or subsidiary apotheosis,
after which He would submit to the experience of sleep. The Aziz
family did not grasp as much as this, being Moslem, but the visit to
the Jail was common knowledge. Smiling, with downcast eyes, the
prisoners discussed with the gentry their chances of salvation.
Except for the irons on their legs, they resembled other men, nor did
they feel different. Five of them, who had not yet been brought to
trial, could expect no pardon, but all who had been convicted were
full of hope. They did not distinguish between the God and the
Rajah in their minds, both were too far above them; but the guard
was better educated, and ventured to enquire after His Highness’s
health.

“It always improves,” replied the medicine man. As a matter of

fact, the Rajah was dead, the ceremony overnight had overtaxed his
strength. His death was being concealed lest the glory of the festival
were dimmed. The Hindu physician, the Private Secretary, and a
confidential servant remained with the corpse, while Aziz had
assumed the duty of being seen in public, and misleading people. He
had liked the ruler very much, and might not prosper under his
successor, yet he could not worry over such problems yet, for he was
involved in the illusion he helped to create. The children continued
to run about, hunting for a frog to put in Mohammed Latif’s bed, the
little fools. Hundreds of frogs lived in their own garden, but they
must needs catch one up on the fort. They reported two topis below.
Fielding and his brother-in-law, instead of resting after their journey,
were climbing the slope to the saint’s tomb!

“Throw stones?” asked Karim.

“Put powdered glass in their pan?”

“Ahmed, come here for such wickedness.” He raised his hand to

smite his firstborn, but allowed it to be kissed instead. It was sweet
to have his sons with him at this moment, and to know they were
affectionate and brave. He pointed out that the Englishmen were
State guests, so must not be poisoned, and received, as always,
gentle yet enthusiastic assent to his words.

The two visitors entered the octagon, but rushed out at once
pursued by some bees. Hither and thither they ran, beating their
heads; the children shrieked with derision, and out of heaven, as if a
plug had been pulled, fell a jolly dollop of rain. Aziz had not meant
to greet his former friend, but the incident put him into an excellent
temper. He felt compact and strong. He shouted out, “Hullo,
gentlemen, are you in trouble?”

The brother-in-law exclaimed; a bee had got him.

“Lie down in a pool of water, my dear sir—here are plenty. Don’t

come near me. . . . I cannot control them, they are State bees;
complain to His Highness of their behaviour.” There was no real
danger, for the rain was increasing. The swarm retired to the shrine.
He went up to the stranger and pulled a couple of stings out of his
wrist, remarking, “Come, pull yourself together and be a man.”

“How do you do, Aziz, after all this time? I heard you were settled
in here,” Fielding called to him, but not in friendly tones. “I suppose
a couple of stings don’t signify.”
 “Not the least. I’ll send an embrocation over to the Guest House. I
heard you were settled in there.”

“Why have you not answered my letters?” he asked, going straight

for the point, but not reaching it, owing to buckets of rain. His
companion, new to the country, cried, as the drops drummed on his
topi, that the bees were renewing their attack. Fielding checked his
antics rather sharply, then said: “Is there a short cut down to our
carriage? We must give up our walk. The weather’s pestilential.”

“Yes. That way.”

“Are you not coming down yourself?”

Aziz sketched a comic salaam; like all Indians, he was skilful in the
slighter impertinences. “I tremble, I obey,” the gesture said, and it
was not lost upon Fielding. They walked down a rough path to the
road—the two men first; the brother-in-law (boy rather than man)
next, in a state over his arm, which hurt; the three Indian children
last, noisy and impudent—all six wet through.

“How goes it, Aziz?”

“In my usual health.”

“Are you making anything out of your life here?”

“How much do you make out of yours?”

“Who is in charge of the Guest House?” he asked, giving up his

slight effort to recapture their intimacy, and growing more official;
he was older and sterner.

“His Highness’s Private Secretary, probably.”

“Where is he, then?”

“I don’t know.”
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.

Let us accompany you on the journey of exploring knowledge and

personal growth!

ebookfinal.com