Introduction to Discrepancy Theory
|
|
- David Oswin Wheeler
- 6 years ago
- Views:
Transcription
1 Introduction to Discrepancy Theory Aleksandar (Sasho) Nikolov University of Toronto Sasho Nikolov (U of T) Discrepancy 1 / 23
2 Outline 1 Monte Carlo 2 Discrepancy and Quasi MC 3 Combinatorial Discrepancy Sasho Nikolov (U of T) Discrepancy 2 / 23
3 Discrepancy Theory How well can a discrete object approximation a continuous object? How well can a small object approximation a big object? Sasho Nikolov (U of T) Discrepancy 3 / 23
4 How to Compute an Integral? A fundamental problem in sciences: How to approximate the integral of a function? 1 0 f (x)dx =? f(x) 1 0 f(x)dx 0 1 Sasho Nikolov (U of T) Discrepancy 4 / 23
5 The Issues Didn t we learn this in calculus? xe x dx = [xe x ] 1 0 e x dx = [(x 1)e x ] Sasho Nikolov (U of T) Discrepancy 5 / 23
6 The Issues Didn t we learn this in calculus? xe x dx = [xe x ] 1 0 e x dx = [(x 1)e x ] Integration is hard! Many interesting functions do not have a closed form integral at all. The function f may be very complicated! Sasho Nikolov (U of T) Discrepancy 5 / 23
7 The Issues Continue Or we may not even know what f really is! f (x) may be: The speed of a particle at time x The price of a stock at time x The amount of energy released when we burn x grams of some chemical substance Sasho Nikolov (U of T) Discrepancy 6 / 23
8 The Issues Continue Or we may not even know what f really is! f (x) may be: The speed of a particle at time x The price of a stock at time x The amount of energy released when we burn x grams of some chemical substance Sometimes all we can do is treat f as a black box: we can query f (x) at a specific x each query may require a new experiment: expensive! Sasho Nikolov (U of T) Discrepancy 6 / 23
9 The Issues Continue Or we may not even know what f really is! f (x) may be: The speed of a particle at time x The price of a stock at time x The amount of energy released when we burn x grams of some chemical substance Sometimes all we can do is treat f as a black box: we can query f (x) at a specific x each query may require a new experiment: expensive! Can we compute 1 0 f (x)dx with a black box f, under minimal assumptions, with few queries? Sasho Nikolov (U of T) Discrepancy 6 / 23
10 The Monte Carlo Idea During the Manhattan Project, von Neumann and Ulam had an idea (inspired by Ulam s uncle s gambling habbit): 1 Pick n random points in [0, 1] 2 Estimate integral by average 0 1 Sasho Nikolov (U of T) Discrepancy 7 / 23
11 The Monte Carlo Idea During the Manhattan Project, von Neumann and Ulam had an idea (inspired by Ulam s uncle s gambling habbit): 1 Pick n random points in [0, 1] 2 Estimate integral by average f(x 1 )f(x 2 ) f(x3 ) f(x 12 ) 0 x 1 x 2 x 3 x 12 1 Sasho Nikolov (U of T) Discrepancy 7 / 23
12 The Monte Carlo Idea During the Manhattan Project, von Neumann and Ulam had an idea (inspired by Ulam s uncle s gambling habbit): 1 Pick n random points in [0, 1] 2 Estimate integral by average f(x 1 )f(x 2 ) f(x3 ) f(x 12 ) 1 f(x)dx i=1 f(x i) 0 x 1 x 2 x 3 x 12 1 Sasho Nikolov (U of T) Discrepancy 7 / 23
13 Monte Carlo: Convergence If we pick n random points x 1,..., x n [0, 1] then 1 f (x)dx 1 n f (x i ) n E(f ), n 0 i=1 where E(f ) is a measure of the energy of f. ( 1 E(f ) = 0 ) 1/2 f (x) 2 dx Sasho Nikolov (U of T) Discrepancy 8 / 23
14 Can we do better? We have a sequence x = (x 1, x 2, x 3,...) that we want to use to estimate integrals, using an average. Sasho Nikolov (U of T) Discrepancy 9 / 23
15 Can we do better? We have a sequence x = (x 1, x 2, x 3,...) that we want to use to estimate integrals, using an average. When we want a more accurate estimate, we evaluate f at a few more points from the sequence. Sasho Nikolov (U of T) Discrepancy 9 / 23
16 Can we do better? We have a sequence x = (x 1, x 2, x 3,...) that we want to use to estimate integrals, using an average. When we want a more accurate estimate, we evaluate f at a few more points from the sequence. We want the error Err(f, x, n) := to be as small as possible. 1 0 f (x)dx 1 n n f (x i ) We know that if x is random, Err(f, x, n) 1/ n? for f with constant energy. i=1 Sasho Nikolov (U of T) Discrepancy 9 / 23
17 Can we do better? We have a sequence x = (x 1, x 2, x 3,...) that we want to use to estimate integrals, using an average. When we want a more accurate estimate, we evaluate f at a few more points from the sequence. We want the error Err(f, x, n) := to be as small as possible. 1 0 f (x)dx 1 n n f (x i ) We know that if x is random, Err(f, x, n) 1/ n? for f with constant energy. i=1 Can we achieve Err(f, x, n) 1/n for all nice f? Sasho Nikolov (U of T) Discrepancy 9 / 23
18 Outline 1 Monte Carlo 2 Discrepancy and Quasi MC 3 Combinatorial Discrepancy Sasho Nikolov (U of T) Discrepancy 10 / 23
19 Intervals Are Enough If a sequence x has small error for all intervals, then it has small error for all smooth functions. δ( x, n) = max a b 1 n {i : a x i b}. a,b [0,1] Sasho Nikolov (U of T) Discrepancy 11 / 23
20 Intervals Are Enough If a sequence x has small error for all intervals, then it has small error for all smooth functions. δ( x, n) = max a b 1 n {i : a x i b}. a,b [0,1] Koksma-Hlawka inequality: Err(f, x, n) V (f )δ( x, n), where V (f ) is a measure of the smoothness of f (total variation). Sasho Nikolov (U of T) Discrepancy 11 / 23
21 Intervals Are Enough If a sequence x has small error for all intervals, then it has small error for all smooth functions. δ( x, n) = max a b 1 n {i : a x i b}. a,b [0,1] Koksma-Hlawka inequality: Err(f, x, n) V (f )δ( x, n), where V (f ) is a measure of the smoothness of f (total variation). van der Corput (1934): Can δ( x, n) = O(1/n) for some sequence x? Sasho Nikolov (U of T) Discrepancy 11 / 23
22 From Intervals to Rectangles Roth showed that studying δ( x, n) is equivalent to placing n points uniformly in a unit square. (Think of one dimension as the index.) n = 20 Sasho Nikolov (U of T) Discrepancy 12 / 23
23 From Intervals to Rectangles Roth showed that studying δ( x, n) is equivalent to placing n points uniformly in a unit square. (Think of one dimension as the index.) n = R 1 2 area(r) - 1 n #{points in R} = = 1 10 Sasho Nikolov (U of T) Discrepancy 12 / 23
24 Discrepancy of Rectangles For a set P of n points in [0, 1] 2 and a rectangle R = [a, b] [c, d] P R d(p, R) = area(r) n = (b a)(d c) d(p) = max D(P, R) R P R n Sasho Nikolov (U of T) Discrepancy 13 / 23
25 Discrepancy of Rectangles For a set P of n points in [0, 1] 2 and a rectangle R = [a, b] [c, d] P R d(p, R) = area(r) n = (b a)(d c) d(p) = max D(P, R) R P R n We can construct a sequence x with δ( x, n) = O(f (n)). For any n, we can construct a set P of n points s.t. d(p) = O(f (n)). Sasho Nikolov (U of T) Discrepancy 13 / 23
26 Grid Can we construct P s.t. d(p) = O(1/n)? Sasho Nikolov (U of T) Discrepancy 14 / 23
27 Grid Can we construct P s.t. d(p) = O(1/n)? Grid: d(p) 1 n : area(r) 0 and P R = n Sasho Nikolov (U of T) Discrepancy 14 / 23
28 Irrational Lattice P = { (i/n, {i 2}) : i = 0,..., n 1 } : d(p) = Θ( log n n ) {x} = fracional part of x = x x Sasho Nikolov (U of T) Discrepancy 15 / 23
29 van der Corput set P = {(i/n, rev(i)) : i = 0,..., n 1}: D(P) = Θ( log n n ) rev(b k b k 1... b 1 b 0 ) = 0.b 1 b 2... b k Sasho Nikolov (U of T) Discrepancy 16 / 23
30 Roth s Lower Bound, and Questions D(P) = O( log n n ) is possible we can estimate integrals with error O But what about d(p) = O(1/n)? ( ) log n n Sasho Nikolov (U of T) Discrepancy 17 / 23
31 Roth s Lower Bound, and Questions D(P) = O( log n n ) is possible we can estimate integrals with error O But what about d(p) = O(1/n)? Theorem (Roth, 1954; Schmidt 1972) For any n-point set P, D(P) = Ω( log n n ). ( ) log n n Sasho Nikolov (U of T) Discrepancy 17 / 23
32 Roth s Lower Bound, and Questions D(P) = O( log n n ) is possible we can estimate integrals with error O But what about d(p) = O(1/n)? Theorem (Roth, 1954; Schmidt 1972) For any n-point set P, D(P) = Ω( log n n ). ( ) log n n What about boxes in dimension 3? In dimension k? for η k 0 as k. (log n) (k 1)/2+η k n d(p) (log n)k 1 n Sasho Nikolov (U of T) Discrepancy 17 / 23
33 Outline 1 Monte Carlo 2 Discrepancy and Quasi MC 3 Combinatorial Discrepancy Sasho Nikolov (U of T) Discrepancy 18 / 23
34 Tusnády s Problem Given: Set Q of n points in the unit square Goal: Color each point p Q red or blue so that each rectangle R is as balanced as possible. Sasho Nikolov (U of T) Discrepancy 19 / 23
35 Tusnády s Problem Given: Set Q of n points in the unit square Goal: Color each point p Q red or blue so that each rectangle R is as balanced as possible. Sasho Nikolov (U of T) Discrepancy 19 / 23
36 Tusnády s Problem Given: Set Q of n points in the unit square Goal: Color each point p Q red or blue so that each rectangle R is as balanced as possible. disc(q) := min χ where χ: P { 1, 1} is a coloring. max R p R P χ(p), Sasho Nikolov (U of T) Discrepancy 19 / 23
37 From Combinatorial to Geometric Discrepancy For any n there exists an n-point set P s.t. d(p) 1 n max disc(q), Q where Q ranges over n-point sets in the unit square. Sasho Nikolov (U of T) Discrepancy 20 / 23
38 From Combinatorial to Geometric Discrepancy For any n there exists an n-point set P s.t. d(p) 1 n max disc(q), Q where Q ranges over n-point sets in the unit square. Sasho Nikolov (U of T) Discrepancy 20 / 23
39 From Combinatorial to Geometric Discrepancy For any n there exists an n-point set P s.t. d(p) 1 n max disc(q), Q where Q ranges over n-point sets in the unit square. Sasho Nikolov (U of T) Discrepancy 20 / 23
40 From Combinatorial to Geometric Discrepancy For any n there exists an n-point set P s.t. d(p) 1 n max disc(q), Q where Q ranges over n-point sets in the unit square. Sasho Nikolov (U of T) Discrepancy 20 / 23
41 From Combinatorial to Geometric Discrepancy For any n there exists an n-point set P s.t. d(p) 1 n max disc(q), Q where Q ranges over n-point sets in the unit square. Sasho Nikolov (U of T) Discrepancy 20 / 23
42 Bounds for Tusnády Theorem (Nikolov, Matoušek, Talwar, 2014) log(n) d 1 max disc(q) log(n)d+1/2 Q The proof uses (the analysis of) an algorithm to estimate discrepancy. Sasho Nikolov (U of T) Discrepancy 21 / 23
43 Computational Questions How can we efficiently (i.e. fast) find balanced colorings? Can we compute disc(q)? This kind of balanced colorings problem has many other applications: computational geometry data structures approximation algorithms private data analysis. Sasho Nikolov (U of T) Discrepancy 22 / 23
44 Sasho Nikolov (U of T) Discrepancy 23 / 23
The Discrepancy Function and the Small Ball Inequality in Higher Dimensions
The Discrepancy Function and the Small Ball Inequality in Higher Dimensions Dmitriy Georgia Institute of Technology (joint work with M. Lacey and A. Vagharshakyan) 2007 Fall AMS Western Section Meeting
More informationMathematical Methods for Neurosciences. ENS - Master MVA Paris 6 - Master Maths-Bio ( )
Mathematical Methods for Neurosciences. ENS - Master MVA Paris 6 - Master Maths-Bio (2014-2015) Etienne Tanré - Olivier Faugeras INRIA - Team Tosca October 22nd, 2014 E. Tanré (INRIA - Team Tosca) Mathematical
More informationLow-discrepancy constructions in the triangle
Low-discrepancy constructions in the triangle Department of Statistics Stanford University Joint work with Art B. Owen. ICERM 29th October 2014 Overview Introduction 1 Introduction The Problem Motivation
More informationDiscrepancy and Uniform Distribution. by Daniel Baczkowski
Discrepancy and Uniform Distribution by Daniel Baczkowski A sequence (u n ) R is uniformly distributed mod 1 if lim N # { n N : {u n } [a, b] } N = b a for every 0 a < b 1. Weyl s Criterion says TFAE 1.
More informationNumerical Methods in Economics MIT Press, Chapter 9 Notes Quasi-Monte Carlo Methods. Kenneth L. Judd Hoover Institution.
1 Numerical Methods in Economics MIT Press, 1998 Chapter 9 Notes Quasi-Monte Carlo Methods Kenneth L. Judd Hoover Institution October 26, 2002 2 Quasi-Monte Carlo Methods Observation: MC uses random sequences
More informationFast integration using quasi-random numbers. J.Bossert, M.Feindt, U.Kerzel University of Karlsruhe ACAT 05
Fast integration using quasi-random numbers J.Bossert, M.Feindt, U.Kerzel University of Karlsruhe ACAT 05 Outline Numerical integration Discrepancy The Koksma-Hlawka inequality Pseudo- and quasi-random
More informationA LeVeque-type lower bound for discrepancy
reprinted from Monte Carlo and Quasi-Monte Carlo Methods 998, H. Niederreiter and J. Spanier, eds., Springer-Verlag, 000, pp. 448-458. A LeVeque-type lower bound for discrepancy Francis Edward Su Department
More informationMonte Carlo Integration
Monte Carlo Integration SCX5005 Simulação de Sistemas Complexos II Marcelo S. Lauretto www.each.usp.br/lauretto Reference: Robert CP, Casella G. Introducing Monte Carlo Methods with R. Springer, 2010.
More informationTUSNÁDY S PROBLEM, THE TRANSFERENCE PRINCIPLE, AND NON-UNIFORM QMC SAMPLING
TUSNÁDY S PROBLEM, THE TRANSFERENCE PRINCIPLE, AND NON-UNIFORM QMC SAMPLING CHRISTOPH AISTLEITNER, DMITRIY BILYK, AND ALEKSANDAR NIKOLOV Abstract. It is well-known that for every N 1 and d 1 there exist
More informationAdvanced Monte Carlo Methods - Computational Challenges
Advanced Monte Carlo Methods - Computational Challenges Ivan Dimov IICT, Bulgarian Academy of Sciences Sofia, February, 2012 Aims To cover important topics from probability theory and statistics. To present
More informationMonte Carlo programs
Monte Carlo programs Alexander Khanov PHYS6260: Experimental Methods is HEP Oklahoma State University November 13, 2017 Monte Carlo methods: the idea In classical physics, if we need to measure something
More informationBayesian Estimation with Sparse Grids
Bayesian Estimation with Sparse Grids Kenneth L. Judd and Thomas M. Mertens Institute on Computational Economics August 7, 27 / 48 Outline Introduction 2 Sparse grids Construction Integration with sparse
More informationQuasi- Monte Carlo Multiple Integration
Chapter 6 Quasi- Monte Carlo Multiple Integration Introduction In some sense, this chapter fits within Chapter 4 on variance reduction; in some sense it is stratification run wild. Quasi-Monte Carlo methods
More informationThe rotor-router mechanism for quasirandom walk and aggregation. Jim Propp (U. Mass. Lowell) July 10, 2008
The rotor-router mechanism for quasirandom walk and aggregation Jim Propp (U. Mass. Lowell) July 10, 2008 (based on articles in progress with Ander Holroyd and Lionel Levine; with thanks also to Hal Canary,
More informationTo: Professor Paul McCudden 9/3/15 Dr. John Sepikas
To: Professor Paul McCudden 9/3/5 Dr. John Sepikas From: Tannaz Oskui Subject: Final Report -Supercomputer Numerical Simulation of Astrophysical Processes JPL/205 CURE My 205 summer CURE project was to
More informationMonte Carlo Methods. Geoff Gordon February 9, 2006
Monte Carlo Methods Geoff Gordon ggordon@cs.cmu.edu February 9, 2006 Numerical integration problem 5 4 3 f(x,y) 2 1 1 0 0.5 0 X 0.5 1 1 0.8 0.6 0.4 Y 0.2 0 0.2 0.4 0.6 0.8 1 x X f(x)dx Used for: function
More informationPseudorandom Sequences II: Exponential Sums and Uniform Distribution
Pseudorandom Sequences II: Exponential Sums and Uniform Distribution Arne Winterhof Austrian Academy of Sciences Johann Radon Institute for Computational and Applied Mathematics Linz Carleton University
More informationApplied Statistics. Monte Carlo Simulations. Troels C. Petersen (NBI) Statistics is merely a quantisation of common sense
Applied Statistics Monte Carlo Simulations Troels C. Petersen (NBI) Statistics is merely a quantisation of common sense 1 Random numbers Any one who considers arithmetical methods of producing random digits
More informationHigh-dimensional permutations and discrepancy
High-dimensional permutations and discrepancy BIRS Meeting, August 2016 What are high dimensional permutations? What are high dimensional permutations? A permutation can be encoded by means of a permutation
More informationChapter 6: The Definite Integral
Name: Date: Period: AP Calc AB Mr. Mellina Chapter 6: The Definite Integral v v Sections: v 6.1 Estimating with Finite Sums v 6.5 Trapezoidal Rule v 6.2 Definite Integrals 6.3 Definite Integrals and Antiderivatives
More informationQMC methods in quantitative finance. and perspectives
, tradition and perspectives Johannes Kepler University Linz (JKU) WU Research Seminar What is the content of the talk Valuation of financial derivatives in stochastic market models using (QMC-)simulation
More informationMonte Carlo Integration
Monte Carlo Integration Ronald Kleiss 1 IMAPP Radboud University, Nijmegen, the Netherlands NIKHEF Academic lectures, 7/12/7-1/12/7 1 R.Kleiss@science.ru.nl 1 1 Ideology Any numerical procedure in which
More informationSolution: First we need to find the mean of this distribution. The mean is. ) = e[( e 1 e 1 ) 0] = 2.
Math 0A with Professor Stankova Worksheet, Discussion #; Monday, //207 GSI name: Roy Zhao Standard Deviation Example. Let f(x) = e e x for x and 0 otherwise. Find the standard deviation of this distribution.
More information1. Let A R be a nonempty set that is bounded from above, and let a be the least upper bound of A. Show that there exists a sequence {a n } n N
Applied Analysis prelim July 15, 216, with solutions Solve 4 of the problems 1-5 and 2 of the problems 6-8. We will only grade the first 4 problems attempted from1-5 and the first 2 attempted from problems
More informationMatrix approximation and Tusnády s problem
European Journal of Combinatorics 28 (2007) 990 995 www.elsevier.com/locate/ejc Matrix approximation and Tusnády s problem Benjamin Doerr Max-Planck-Institut für Informatik, Stuhlsatzenhausweg 85, 66123
More informationHigh Dimensional Integration: New Weapons Fighting the Curse of Dimensionality
High Dimensional Integration: ew Weapons Fighting the Curse of Dimensionality Peter Zinterhof Technical Report 26-4 October 26 Department of Computer Sciences Jakob-Haringer-Straße 2 52 Salzburg Austria
More informationThe Growth of Functions and Big-O Notation
The Growth of Functions and Big-O Notation Big-O Notation Big-O notation allows us to describe the aymptotic growth of a function without concern for i) constant multiplicative factors, and ii) lower-order
More informationFrom Random Numbers to Monte Carlo. Random Numbers, Random Walks, Diffusion, Monte Carlo integration, and all that
From Random Numbers to Monte Carlo Random Numbers, Random Walks, Diffusion, Monte Carlo integration, and all that Random Walk Through Life Random Walk Through Life If you flip the coin 5 times you will
More informationLesson 29 MA Nick Egbert
Lesson 9 MA 16 Nick Egbert Overview In this lesson we build on the previous two b complicating our domains of integration and discussing the average value of functions of two variables. Lesson So far the
More informationNEW SEQUENCES FROM OLD: A SEARCH FOR LOW DISCREPANCY SEQUENCES. Jarvis Haupt
EW SEQUECES FROM OLD: A SEARCH FOR LOW DISCREPACY SEQUECES Jarvis Haupt Electrical and Computer Engineering Department University of Wisconsin, Madison, WI, USA jdhaupt@wisc.edu ABSTRACT Aside from their
More information4. Numerical Quadrature. Where analytical abilities end... continued
4. Numerical Quadrature Where analytical abilities end... continued Where analytical abilities end... continued, November 30, 22 1 4.3. Extrapolation Increasing the Order Using Linear Combinations Once
More informationUniformly Distributed Sequences
Uniformly Distributed Sequences Arman Tavakoli University of British Columbia July 13,2012 Motivation - Monte-Carlo Method Example Estimating π: Pick a random point from the square, the probability of
More informationMATH 2413 TEST ON CHAPTER 4 ANSWER ALL QUESTIONS. TIME 1.5 HRS.
MATH 1 TEST ON CHAPTER ANSWER ALL QUESTIONS. TIME 1. HRS. M1c Multiple Choice Identify the choice that best completes the statement or answers the question. 1. Use the summation formulas to rewrite the
More informationChapter 4: Monte-Carlo Methods
Chapter 4: Monte-Carlo Methods A Monte-Carlo method is a technique for the numerical realization of a stochastic process by means of normally distributed random variables. In financial mathematics, it
More informationAPPLICATIONS OF INTEGRATION
6 APPLICATIONS OF INTEGRATION APPLICATIONS OF INTEGRATION 6.5 Average Value of a Function In this section, we will learn about: Applying integration to find out the average value of a function. AVERAGE
More informationGAUSS S CIRCLE PROBLEM
GAUSS S CIRCLE PROBLEM UPDATE: These are old course notes which are no longer being revised. For the most recent version please see my pre-book at www.math.uga.edu/ pete/4400full.pdf, in which this handout
More informationQMC3: QMC beyond the cube 1. QMC beyond the cube. Art B. Owen, Stanford University based on joint work with Kinjal Basu, LinkedIn
QMC3: QMC beyond the cube 1 QMC beyond the cube Art B. Owen, Stanford University based on joint work with Kinjal Basu, LinkedIn QMC3: QMC beyond the cube 2 Kinjal Basu PhD 2016, Stanford Statistics QMC3:
More informationMATH 18.01, FALL PROBLEM SET #5 SOLUTIONS (PART II)
MATH 8, FALL 7 - PROBLEM SET #5 SOLUTIONS (PART II (Oct ; Antiderivatives; + + 3 7 points Recall that in pset 3A, you showed that (d/dx tanh x x Here, tanh (x denotes the inverse to the hyperbolic tangent
More informationLow-Discrepancy Curves and Efficient Coverage of Space
Low-Discrepancy Curves and Efficient Coverage of Space Subramanian Ramamoorthy 1,3, Ram Rajagopal 2, Qing Ruan 3, and Lothar Wenzel 3 1 Department of Electrical and Computer Engineering, The University
More informationAssignment 9 Mathematics 2(Model Answer)
Assignment 9 Mathematics (Model Answer) The Integral and Comparison Tests Problem: Determine converges or divergence of the series. ) (a) 0 = (b) ) (a) =8 (b) + 3) (a) = (b) 3 + ) (a) e = (b) 5) (a) =0
More informationLecture 15 Random variables
Lecture 15 Random variables Weinan E 1,2 and Tiejun Li 2 1 Department of Mathematics, Princeton University, weinan@princeton.edu 2 School of Mathematical Sciences, Peking University, tieli@pku.edu.cn No.1
More informationNumerical integration - I. M. Peressi - UniTS - Laurea Magistrale in Physics Laboratory of Computational Physics - Unit V
Numerical integration - I M. Peressi - UniTS - Laurea Magistrale in Physics Laboratory of Computational Physics - Unit V deterministic methods in 1D equispaced points (trapezoidal, Simpson...), others...
More informationMath 122 Fall Unit Test 1 Review Problems Set A
Math Fall 8 Unit Test Review Problems Set A We have chosen these problems because we think that they are representative of many of the mathematical concepts that we have studied. There is no guarantee
More informationQUASI-RANDOM NUMBER GENERATION IN THE VIEW OF PARALLEL IMPLEMENTATION. 1. Introduction
Trends in Mathematics Information Center for Mathematical Sciences Volume 6, Number 2, December, 2003, Pages 155 162 QUASI-RANDOM NUMBER GENERATION IN THE VIEW OF PARALLEL IMPLEMENTATION CHI-OK HWANG Abstract.
More informationPart III. Quasi Monte Carlo methods 146/349
Part III Quasi Monte Carlo methods 46/349 Outline Quasi Monte Carlo methods 47/349 Quasi Monte Carlo methods Let y = (y,...,y N ) be a vector of independent uniform random variables in Γ = [0,] N, u(y)
More informationDay 2 Notes: Riemann Sums In calculus, the result of f ( x)
AP Calculus Unit 6 Basic Integration & Applications Day 2 Notes: Riemann Sums In calculus, the result of f ( x) dx is a function that represents the anti-derivative of the function f(x). This is also sometimes
More informationAMath 483/583 Lecture 26. Notes: Notes: Announcements. Notes: AMath 483/583 Lecture 26. Outline:
AMath 483/583 Lecture 26 Outline: Monte Carlo methods Random number generators Monte Carlo integrators Random walk solution of Poisson problem Announcements Part of Final Project will be available tomorrow.
More informationCOURSE Numerical integration of functions (continuation) 3.3. The Romberg s iterative generation method
COURSE 7 3. Numerical integration of functions (continuation) 3.3. The Romberg s iterative generation method The presence of derivatives in the remainder difficulties in applicability to practical problems
More informationHigh-dimensional Problems in Finance and Economics. Thomas M. Mertens
High-dimensional Problems in Finance and Economics Thomas M. Mertens NYU Stern Risk Economics Lab April 17, 2012 1 / 78 Motivation Many problems in finance and economics are high dimensional. Dynamic Optimization:
More informationReview: Basic MC Estimator
Overview Earlier lectures Monte Carlo I: Integration Signal processing and sampling Today: Monte Carlo II Noise and variance Efficiency Stratified sampling Importance sampling Review: Basic MC Estimator
More information2.8 Linear Approximation and Differentials
2.8 Linear Approximation Contemporary Calculus 1 2.8 Linear Approximation and Differentials Newton's method used tangent lines to "point toward" a root of the function. In this section we examine and use
More informationOBJECTIVES Use the area under a graph to find total cost. Use rectangles to approximate the area under a graph.
4.1 The Area under a Graph OBJECTIVES Use the area under a graph to find total cost. Use rectangles to approximate the area under a graph. 4.1 The Area Under a Graph Riemann Sums (continued): In the following
More informationTutorial on quasi-monte Carlo methods
Tutorial on quasi-monte Carlo methods Josef Dick School of Mathematics and Statistics, UNSW, Sydney, Australia josef.dick@unsw.edu.au Comparison: MCMC, MC, QMC Roughly speaking: Markov chain Monte Carlo
More informationPROBABILITY VITTORIA SILVESTRI
PROBABILITY VITTORIA SILVESTRI Contents Preface. Introduction 2 2. Combinatorial analysis 5 3. Stirling s formula 8 4. Properties of Probability measures Preface These lecture notes are for the course
More informationNew approaches to uncertainty evaluation:
New approaches to uncertainty evaluation: Propagation of uncertainty, numerical methods and Monte Carlo simulation Wolfram Bremser BAM Federal Institute for Materials Research and Testing Unter den Eichen
More informationThe integral test and estimates of sums
The integral test Suppose f is a continuous, positive, decreasing function on [, ) and let a n = f (n). Then the series n= a n is convergent if and only if the improper integral f (x)dx is convergent.
More informationScience One Integral Calculus
Science One Integral Calculus January 018 Happy New Year! Differential Calculus central idea: The Derivative What is the derivative f (x) of a function f(x)? Differential Calculus central idea: The Derivative
More informationMath Refresher Course
Math Refresher Course Columbia University Department of Political Science Fall 2007 Day 2 Prepared by Jessamyn Blau 6 Calculus CONT D 6.9 Antiderivatives and Integration Integration is the reverse of differentiation.
More informationΩ = e E d {0, 1} θ(p) = P( C = ) So θ(p) is the probability that the origin belongs to an infinite cluster. It is trivial that.
2 Percolation There is a vast literature on percolation. For the reader who wants more than we give here, there is an entire book: Percolation, by Geoffrey Grimmett. A good account of the recent spectacular
More informationMarkov chain Monte Carlo
Markov chain Monte Carlo Peter Beerli October 10, 2005 [this chapter is highly influenced by chapter 1 in Markov chain Monte Carlo in Practice, eds Gilks W. R. et al. Chapman and Hall/CRC, 1996] 1 Short
More informationQuasi-Monte Carlo Algorithms with Applications in Numerical Analysis and Finance
Quasi-Monte Carlo Algorithms with Applications in Numerical Analysis and Finance Reinhold F. Kainhofer Rigorosumsvortrag, Institut für Mathematik Graz, 16. Mai 2003 Inhalt 1. Quasi-Monte Carlo Methoden
More informationAn Algorithmist s Toolkit Nov. 10, Lecture 17
8.409 An Algorithmist s Toolkit Nov. 0, 009 Lecturer: Jonathan Kelner Lecture 7 Johnson-Lindenstrauss Theorem. Recap We first recap a theorem (isoperimetric inequality) and a lemma (concentration) from
More informationTractability of Multivariate Problems
Erich Novak University of Jena Chemnitz, Summer School 2010 1 Plan for the talk Example: approximation of C -functions What is tractability? Tractability by smoothness? Tractability by sparsity, finite
More informationd(x n, x) d(x n, x nk ) + d(x nk, x) where we chose any fixed k > N
Problem 1. Let f : A R R have the property that for every x A, there exists ɛ > 0 such that f(t) > ɛ if t (x ɛ, x + ɛ) A. If the set A is compact, prove there exists c > 0 such that f(x) > c for all x
More informationComputer intensive statistical methods Lecture 1
Computer intensive statistical methods Lecture 1 Jonas Wallin Chalmers, Gothenburg. Jonas Wallin - jonwal@chalmers.se 1/27 People and literature The following people are involved in the course: Function
More informationChange Of Variable Theorem: Multiple Dimensions
Change Of Variable Theorem: Multiple Dimensions Moulinath Banerjee University of Michigan August 30, 01 Let (X, Y ) be a two-dimensional continuous random vector. Thus P (X = x, Y = y) = 0 for all (x,
More informationSCRAMBLED GEOMETRIC NET INTEGRATION OVER GENERAL PRODUCT SPACES. Kinjal Basu Art B. Owen
SCRAMBLED GEOMETRIC NET INTEGRATION OVER GENERAL PRODUCT SPACES By Kinjal Basu Art B. Owen Technical Report No. 2015-07 March 2015 Department of Statistics STANFORD UNIVERSITY Stanford, California 94305-4065
More informationLecture 4 February 2
4-1 EECS 281B / STAT 241B: Advanced Topics in Statistical Learning Spring 29 Lecture 4 February 2 Lecturer: Martin Wainwright Scribe: Luqman Hodgkinson Note: These lecture notes are still rough, and have
More information2tdt 1 y = t2 + C y = which implies C = 1 and the solution is y = 1
Lectures - Week 11 General First Order ODEs & Numerical Methods for IVPs In general, nonlinear problems are much more difficult to solve than linear ones. Unfortunately many phenomena exhibit nonlinear
More informationA Violation Of The Area Law For Fermionic Entanglement Entropy
A Violation Of The Area Law For Fermionic Entanglement Entropy or: How much can you say in a phone call? Robert Helling 1 Hajo Leschke 2 and Wolfgang Spitzer 2 1 Arnold Sommerfeld Center Ludwig-Maximilians-Universität
More informationFE610 Stochastic Calculus for Financial Engineers. Stevens Institute of Technology
FE610 Stochastic Calculus for Financial Engineers Lecture 3. Calculaus in Deterministic and Stochastic Environments Steve Yang Stevens Institute of Technology 01/31/2012 Outline 1 Modeling Random Behavior
More informationSection 4.4. Using the Fundamental Theorem. Difference Equations to Differential Equations
Difference Equations to Differential Equations Section 4.4 Using the Fundamental Theorem As we saw in Section 4.3, using the Fundamental Theorem of Integral Calculus reduces the problem of evaluating a
More informationMath 205b Homework 2 Solutions
Math 5b Homework Solutions January 5, 5 Problem (R-S, II.) () For the R case, we just expand the right hand side and use the symmetry of the inner product: ( x y x y ) = = ((x, x) (y, y) (x, y) (y, x)
More information3. Numerical Quadrature. Where analytical abilities end...
3. Numerical Quadrature Where analytical abilities end... Numerisches Programmieren, Hans-Joachim Bungartz page 1 of 32 3.1. Preliminary Remarks The Integration Problem Numerical quadrature denotes numerical
More informationMath 162: Calculus IIA
Math 62: Calculus IIA Final Exam ANSWERS December 9, 26 Part A. (5 points) Evaluate the integral x 4 x 2 dx Substitute x 2 cos θ: x 8 cos dx θ ( 2 sin θ) dθ 4 x 2 2 sin θ 8 cos θ dθ 8 cos 2 θ cos θ dθ
More informationSpring 2015, Math 111 Lab 9: The Definite Integral as the Are. the Area under a Curve
Spring 2015, Math 111 Lab 9: The Definite Integral as the Area under a Curve William and Mary April 14, 2015 Historical Outline Intuition Learning Objectives Today, we will be looking at applications of
More informationDecoupling course outline Decoupling theory is a recent development in Fourier analysis with applications in partial differential equations and
Decoupling course outline Decoupling theory is a recent development in Fourier analysis with applications in partial differential equations and analytic number theory. It studies the interference patterns
More informationFibonacci sets and symmetrization in discrepancy theory
Fibonacci sets and symmetrization in discrepancy theory Dmitriy Bilyk, V.N. Temlyakov, Rui Yu Department of Mathematics, University of South Carolina, Columbia, SC, 29208, U.S.A. Abstract We study the
More informationStatistical Methods in Particle Physics
Statistical Methods in Particle Physics 4. Monte Carlo Methods Prof. Dr. Klaus Reygers (lectures) Dr. Sebastian Neubert (tutorials) Heidelberg University WS 2017/18 Monte Carlo Method Any method which
More informationProgress in high-dimensional numerical integration and its application to stochastic optimization
Progress in high-dimensional numerical integration and its application to stochastic optimization W. Römisch Humboldt-University Berlin Department of Mathematics www.math.hu-berlin.de/~romisch Short course,
More informationA new class of pseudodifferential operators with mixed homogenities
A new class of pseudodifferential operators with mixed homogenities Po-Lam Yung University of Oxford Jan 20, 2014 Introduction Given a smooth distribution of hyperplanes on R N (or more generally on a
More information= (, ) V λ (1) λ λ ( + + ) P = [ ( ), (1)] ( ) ( ) = ( ) ( ) ( 0 ) ( 0 ) = ( 0 ) ( 0 ) 0 ( 0 ) ( ( 0 )) ( ( 0 )) = ( ( 0 )) ( ( 0 )) ( + ( 0 )) ( + ( 0 )) = ( + ( 0 )) ( ( 0 )) P V V V V V P V P V V V
More informationExponential and. Logarithmic Functions. Exponential Functions. Logarithmic Functions
Chapter Five Exponential and Logarithmic Functions Exponential Functions Logarithmic Functions Properties of Logarithms Exponential Equations Exponential Situations Logarithmic Equations Exponential Functions
More informationSOLUTIONS OF SELECTED PROBLEMS
SOLUTIONS OF SELECTED PROBLEMS Problem 36, p. 63 If µ(e n < and χ En f in L, then f is a.e. equal to a characteristic function of a measurable set. Solution: By Corollary.3, there esists a subsequence
More informationUniformly distributed sequences of partitions
Uniformly distributed sequences of partitions Aljoša Volčič Università della Calabria Uniform distribution and QMC methods A. Volčič (Università della Calabria) U.d. sequences of partitions Linz, October
More informationFINAL REVIEW FOR MATH The limit. a n. This definition is useful is when evaluating the limits; for instance, to show
FINAL REVIEW FOR MATH 500 SHUANGLIN SHAO. The it Define a n = A: For any ε > 0, there exists N N such that for any n N, a n A < ε. This definition is useful is when evaluating the its; for instance, to
More informationQUASI-MONTE CARLO METHODS: WHERE RANDOMNESS AND DETERMINISM COLLIDE
QUASI-MONTE CARLO METHODS: WHERE RANDOMNESS AND DETERMINISM COLLIDE Michael Mascagni: mascagni@cs.fsu.edu Home Page: http://www.cs.fsu.edu/~mascagni Department of Computer Science 203 Love Building Florida
More informationRandom and Quasi-Random Point Sets
Peter Hellekalek Gerhard Larcher (Editors) Random and Quasi-Random Point Sets With contributions by JözsefBeck Peter Hellekalek Fred J. Hickemell Gerhard Larcher Pierre L'Ecuyer Harald Niederreiter Shu
More informationAutomatic Differentiation and Neural Networks
Statistical Machine Learning Notes 7 Automatic Differentiation and Neural Networks Instructor: Justin Domke 1 Introduction The name neural network is sometimes used to refer to many things (e.g. Hopfield
More informationExtensible grids: uniform sampling on a space-filling curve
Extensible grids: uniform sampling on a space-filling curve Zhijian He Tsinghua University Art B. Owen Stanford University June 2014 Abstract We study the properties of points in [0, 1] d generated by
More informationIntroduction to Reinforcement Learning. CMPT 882 Mar. 18
Introduction to Reinforcement Learning CMPT 882 Mar. 18 Outline for the week Basic ideas in RL Value functions and value iteration Policy evaluation and policy improvement Model-free RL Monte-Carlo and
More informationSimilar Shapes and Gnomons
Similar Shapes and Gnomons May 12, 2013 1. Similar Shapes For now, we will say two shapes are similar if one shape is a magnified version of another. 1. In the picture below, the square on the left is
More informationMarkov Chain Monte Carlo Inference. Siamak Ravanbakhsh Winter 2018
Graphical Models Markov Chain Monte Carlo Inference Siamak Ravanbakhsh Winter 2018 Learning objectives Markov chains the idea behind Markov Chain Monte Carlo (MCMC) two important examples: Gibbs sampling
More informationAP Calculus Chapter 9: Infinite Series
AP Calculus Chapter 9: Infinite Series 9. Sequences a, a 2, a 3, a 4, a 5,... Sequence: A function whose domain is the set of positive integers n = 2 3 4 a n = a a 2 a 3 a 4 terms of the sequence Begin
More informationMath 456: Mathematical Modeling. Tuesday, April 9th, 2018
Math 456: Mathematical Modeling Tuesday, April 9th, 2018 The Ergodic theorem Tuesday, April 9th, 2018 Today 1. Asymptotic frequency (or: How to use the stationary distribution to estimate the average amount
More informationMonte Carlo Integration. Computer Graphics CMU /15-662, Fall 2016
Monte Carlo Integration Computer Graphics CMU 15-462/15-662, Fall 2016 Talk Announcement Jovan Popovic, Senior Principal Scientist at Adobe Research will be giving a seminar on Character Animator -- Monday
More informationChapter 13: Integral Calculus. SSMth2: Basic Calculus Science and Technology, Engineering and Mathematics (STEM) Strands Mr. Migo M.
Chapter 13: Integral Calculus SSMth2: Basic Calculus Science and Technology, Engineering and Mathematics (STEM) Strands Mr. Migo M. Mendoza Chapter 13: Integral Calculus Lecture 13.1: The Integrals Lecture
More informationNumerical Methods with Lévy Processes
Numerical Methods with Lévy Processes 1 Objective: i) Find models of asset returns, etc ii) Get numbers out of them. Why? VaR and risk management Valuing and hedging derivatives Why not? Usual assumption:
More informationLecture 1.4: Inner products and orthogonality
Lecture 1.4: Inner products and orthogonality Matthew Macauley Department of Mathematical Sciences Clemson University http://www.math.clemson.edu/~macaule/ Math 4340, Advanced Engineering Mathematics M.
More informationModeling Accumulations: Introduction to the Issues 11/07/2011
Modeling Accumulations: Introduction to the Issues /07/0 The purpose of calculus is twofold:. to find how something is changing, given what it s doing;. to find what something is doing, given how it s
More information