By Pascal Massart, Stéphane Boucheron, Gábor Lugosi

Focus inequalities for features of self sufficient random variables is a space of likelihood conception that has witnessed a good revolution within the previous few a long time, and has functions in a large choice of components comparable to desktop studying, records, discrete arithmetic, and high-dimensional geometry. approximately talking, if a functionality of many autonomous random variables doesn't count an excessive amount of on any of the variables then it's focused within the feel that with excessive likelihood, it truly is as regards to its anticipated worth. This ebook deals a number of inequalities to demonstrate this wealthy concept in an obtainable means by means of masking the most important advancements and purposes within the box.

The authors describe the interaction among the probabilistic constitution (independence) and a number of instruments starting from practical inequalities to transportation arguments to info conception. functions to the examine of empirical procedures, random projections, random matrix concept, and threshold phenomena also are awarded.

A self-contained advent to focus inequalities, it contains a survey of focus of sums of self reliant random variables, variance bounds, the entropy procedure, and the transportation approach. Deep connections with isoperimetric difficulties are printed while particular realization is paid to purposes to the supremum of empirical processes.

Written through best specialists within the box and containing broad workout sections this booklet may be a useful source for researchers and graduate scholars in arithmetic, theoretical machine technological know-how, and engineering.


The transparent exposition from simple fabric as much as contemporary refined effects and lucid writing sort make the textual content a excitement to learn. newcomers in addition to skilled scientists will prot both from it. it's going to definitely turn into one of many typical references within the box. Hilmar Mai, Zentralblatt Math

Show description

Read or Download Concentration Inequalities: A Nonasymptotic Theory of Independence PDF

Similar probability books

Introduction to Probability Models (10th Edition)

Ross's vintage bestseller, creation to chance types, has been used largely by means of professors because the basic textual content for a primary undergraduate path in utilized chance. It offers an creation to effortless likelihood idea and stochastic strategies, and indicates how likelihood concept might be utilized to the research of phenomena in fields similar to engineering, laptop technological know-how, administration technology, the actual and social sciences, and operations learn.

Real analysis and probability

This vintage textbook, now reissued, deals a transparent exposition of contemporary likelihood thought and of the interaction among the houses of metric areas and likelihood measures. the recent variation has been made much more self-contained than earlier than; it now incorporates a origin of the genuine quantity process and the Stone-Weierstrass theorem on uniform approximation in algebras of services.

Extra resources for Concentration Inequalities: A Nonasymptotic Theory of Independence

Example text

Xi ), and use the convention E0 = E, then we may define i = Ei Z – Ei–1 Z for every i = 1, . . , n. Starting from the decomposition n Z – EZ = i i=1 one has 2 n Var (Z) = E n = i i=1 Now, if j > i, Ei j +2 E i j . j>i i=1 = 0 implies that Ei and, a fortiori, E of the variance: 2 i E j i j i = i Ei j = 0, = 0. Thus, we obtain the following analog of the additivity formula 54 | BOUNDING THE VARIANCE 2 n Var (Z) = E i n = i=1 2 i E . i=1 Until now, we have made no use of the fact that Z is a function of independent variables X1 , .

The idea of random projections was introduced by Frankl and Maehara (1988, 1990), but see also Gupta and Dasgupta (2002) for a particularly simple proof. Achlioptas (2003) considered projections based on Rademacher random variables. 13 is adapted from the arguments of Achlioptas. Random projections have been used successfully in a variety of applications, for example, Linial, London, and Rabinovich (1995), Kleinberg (1997), and Indyk and Motwani (1998). For a survey we refer the reader to the book of Vempala (2004).

We denote the collection of such random variables by G(v). 16). This definition is natural as we know from the previous section that exp (λ2 v/2) is the moment-generating function of a centered normal random SUB-GAUSSIAN RANDOM VARIABLES | 25 variable Y with variance v. Hence, the above definition says that a centered random variable X belongs to G (v) if the moment-generating function of X is dominated by the momentgenerating function of Y. This notion is also convenient because it is naturally stable under convolution in the sense that if X1 , .

Download PDF sample

Rated 4.87 of 5 – based on 50 votes