Tutorial on quasi-monte Carlo methods
|
|
- Kenneth Franklin
- 6 years ago
- Views:
Transcription
1 Tutorial on quasi-monte Carlo methods Josef Dick School of Mathematics and Statistics, UNSW, Sydney, Australia
2 Comparison: MCMC, MC, QMC Roughly speaking: Markov chain Monte Carlo and quasi-monte Carlo are for different types of problems; If you have a problem where Monte Carlo does not work, then chances are quasi-monte Carlo will not work as well; If Monte Carlo works, but you want a faster method try (randomized) quasi-monte Carlo (some tweaking might be necessary). Quasi-Monte Carlo is an "experimental design" approach to Monte Carlo simulation; In this talk we shall discuss how quasi-monte Carlo can be faster than Monte Carlo under certain assumptions.
3 Outline DISCREPANCY, REPRODUCING KERNEL HILBERT SPACES AND WORST-CASE ERROR QUASI-MONTE CARLO POINT SETS RANDOMIZATIONS WEIGHTED FUNCTION SPACES AND TRACTABILITY
4 The task is to approximate an integral I s (f ) = f (z) dz [,1] s for some integrand f by some quadrature rule Q N,s (f ) = 1 N 1 f (x n ) N n= at some sample points x,..., x N 1 [, 1] s.
5 In other words: f (x) dx 1 N 1 f (x n ) [,1] N s n= Area under curve = Volume of cube average function value. If x,..., x N 1 [, 1] s are chosen randomly Monte Carlo If x,..., x N 1 [, 1] s chosen deterministically Quasi-Monte Carlo
6 Smooth integrands Integrand f : [, 1] R; say continuously differentiable; We want to study the integration error: Representation: f (x) = f (1) 1 1 f (x) dx 1 N 1 f (x n ). N n= f (t) dt = f (1) 1 x 1 [,t] (x)f (t) dt, where 1 [,t] (x) = { 1 if x [, t], otherwise.
7 Integration error 1 Substitute: 1 f (x)dx = 1 ( = f (1) N 1 1 f (x n ) = 1 N N n= Integration error: f (x) dx 1 N 1 f (x n ) = N n= = f (1) 1 f (1) ( N 1 f (1) n= 1 1 [,t] (x)f (t) dt ) 1 [,t] (x)f (t) dx dt, 1 N 1 1 N n= ( N [,t] (x n ) N n= 1 [,t] (x n )f (t) dt 1 [,t] (x n )f (t) dt. 1 1 [,t] (x) dx dx ) ) f (t) dt.
8 Local discrepancy Integration error: 1 f (x) dx 1 N 1 f (x n ) = N n= 1 ( ) N [,t] (x n ) t f (t) dt. N Let P = {x,..., x N 1 } [, 1]. Define the local discrepancy by n= P (t) = 1 N 1 1 [,t] (x n ) t, t [, 1]. N n= Then 1 f (x) dx 1 N 1 f (x n ) = N n= 1 P (t)f (t) dt.
9 Koksma-Hlawka inequality 1 f (x) dx 1 N 1 1 f (x n ) N = P (t)f (t) dt n= ( ) 1/p ( 1 1 P (t) p dt f (t) q dt where g Lp = ( g p) 1/p. = P Lp f Lq, 1 p + 1 q = 1, ) 1/q
10 Interpretation of discrepancy Let P = {x,..., x N 1 } [, 1]. Recall the definition of the local discrepancy: P (t) = 1 N 1 1 [,t] (x n ) t, t [, 1]. N n= Local discrepancy measures difference between uniform distribution and emperical distribution of quadrature points P = {x,..., x N 1 }. This is the Kolmogorov-Smirnov test for the difference between the empirical distribution of {x,..., x N 1 } and the uniform distribution.
11 Function space Representation Define inner product: f (x) = f (1) 1 x f (t) dt. f, g = f (1)g(1) + 1 f (t)g (t) dt. and norm Function space: f = f (1) f (t) 2 dt. H = {f : [, 1] R : f absolutely continuous and f < }.
12 Worst-case error The worst-case error is defined by 1 e(h, P) = sup f (x) dx 1 N 1 f (x n ) N. f H, f 1 n=
13 Reproducing kernel Hilbert space Recall: and f (y) = f (1) f, g = f (1)g(1) + f (x)( 1 [,x] (y)) dx 1 f (x)g (x) dx. Goal: Find set of functions g y H for each y [, 1] such that Conclusions: g y (1) = 1 for all y [, 1]; f, g y = f (y) for all f H. g y(x) = 1 [,x] (y) = Make g continuous such that g H; { 1 if y x, otherwise;
14 Reproducing kernel Hilbert space It follows that The function g y (x) = 1 + min{1 x, 1 y}. K (x, y) := g y (x) = 1 + min{1 x, 1 y}, x, y [, 1] is called reproducing kernel. The space H is called a reproducing kernel Hilbert space (with reproducing kernel K ).
15 Numerical integration in reproducing kernel Hilbert spaces Function representation: 1 f (z) dz = 1 N 1 1 f (x n ) = 1 N N n= f, K (, z) dz = f, 1 K (, z) dz N 1 f, K (, x n ) = f, 1 N 1 K (, x n ) N n= Integration error 1 f (z) dz 1 N 1 f (x n ) = f, N n= 1 = f, h, n= K (, z) dz 1 N 1 K (, x n ) N n= where h(x) = 1 K (x, z) dz 1 N 1 K (x, x n ). N n=
16 Worst-case error in reproducing kernel Hilbert spaces Thus e(h, P) = sup f H, f 1 = sup f H, f =1 = sup = f H,f h, h h 1 f, h h f f, = h, f (z) dz 1 N 1 f (x n ) N since the supremum is attained when choosing f = n= h h H.
17 Worst-case error in reproducing kernel Hilbert spaces e 2 (H, P) = N 1 N 2 n,m= K (x, y) dx dy 2 N K (x n, x m ) N 1 1 n= K (x, x n ) dx
18 Numerical integration in higher dimensions Tensor product space H s = H H. Reproducing kernel K (x, y) = s [1 + min{1 x i, 1 y i }], i=1 where x = (x 1,..., x s ), y = (y 1,..., y s ) [, 1] s. Functions f H s have partial mixed derivatives up to order 1 in each variable u f x u L 2 ([, 1] s ), where u {1,..., s}, x u = (x i ) i u and u denotes the cardinality of u and where u f x u (x u, 1) = for all u {1,..., s}.
19 Worst-case error Again e 2 (H, P) = [,1] s + 1 N 1 N 2 n,m= [,1] s K (x, y) dx dy 2 N K (x n, x m ) N 1 n= [,1] s K (x, x n ) dx and where e 2 (H, P) = P (x) s f (x) dx, [,1] x s P (x) = 1 N 1 1 [,x] (x n ) N n= s x i. i=1
20 Discrepancy in higher dimensions Point set P = {x,..., x N 1 } [, 1] s, t = (t 1,..., t s ) [, 1] s. Local discrepancy: where [, t] = s i=1 [, t i]. P (t) = 1 N 1 1 [,t] (x n ) N n= s t i, i=1
21 Koksma-Hlawka inequality Let f : [, 1] s R with ( f = [,1] s s x f (x) q dx) 1/q, and where u f x u (x u, 1) = for all u {1,..., s}. Then f (x) dx 1 N 1 f (x n ) [,1] N f P Lp, s where 1 p + 1 q = 1. n= Construct points P = {x,..., x N 1 } [, 1] s with small discrepancy L p (P) = P Lp. It is often useful to consider different reproducing kernels yielding different worst-case errors and discrepancies. We will see further examples later.
22 Constructions of low-discrepancy sequences Lattice rules (Bilyk, Brauchart, Cools, D., Hellekalek, Hickernell, Hlawka, Joe, Keller, Korobov, Kritzer, Kuo, Larcher, L Ecuyer, Lemieux, Leobacher, Niederreiter, Nuyens, Pillichshammer, Sinescu, Sloan, Temlyakov, Wang, Woźniakowski,...) Digital nets and sequences (Baldeaux, Bierbrauer, Brauchart, Chen, D., Edel, Faure, Hellekalek, Hofer, Keller, Kritzer, Kuo, Larcher, Leobacher, Niederreiter, Owen, Özbudak, Pillichshammer, Pirsic, Schmid, Skriganov, Sobol, Wang, Xing, Yue,...) Hammersley-Halton sequences (Atanassov, De Clerck, Faure, Halton, Hammersley, Kritzer, Larcher, Lemieux, Pillichshammer, Pirsic, White,...) Kronecker sequences (Beck, Hellekalek, Larcher, Niederreiter, Schoissengeier,...)
23 Lattice rules Let N N, let g = (g 1,..., g s ) {1,..., N 1} s. Choose the quadrature points as { ng } x n =, for n =,..., N 1, N where {z} = z z for z R +.
24 Fibonacci lattice rules Lattice rule with N = 55 points and generating vector g = (1, 34).
25 Fibonacci lattice rules Lattice rule with N = 89 points and generating vector g = (1, 55).
26 In comparison: Random point set Random set of 64 points generated by Matlab using Mersenne Twister.
27 Lattice rules How to find generating vector g? Reproducing kernel: s K (x, y) = (1 + 2πB α ({x i y i })), i=1 where {x i y i } = (x i y i ) x i y i is the fractional part of x i y i. Reproducing kernel Hilbert space of Fourier series: f (x) = f (h)e 2πih x. h Z s
28 Worst-case error: e 2 α(g) = N N 1 s ( n= i= πB α ({ ngi N })), where B α is the Bernoulli polynomial of order α. For instance B 2 (x) = x 2 x + 1/6.
29 Component-by-component construction (Korobov, Sloan-Reztsov,Sloan-Kuo-Joe, Nuyens-Cools) Set g 1 = 1. For d = 2,..., s assume that we have found g 2,..., g d 1. Then find g d {1,..., N 1} which minimizes e α (g 1,..., g d 1, g) as a function of g, i.e. g d = argmin g {1,...,N 1} e2 α(g 1,..., g d 1, g). Using fast Fourier transform (Nuyens, Cools), a good generating vector g {1,..., N 1} s can be found in O(sN log N) operations.
30 Hammersley-Halton sequence Radical inverse function in base b: Let n N have base b expansion n = n + n 1 b + + n a 1 b a 1. Set ϕ b (n) = n b + n 1 b n a 1 b a [, 1]. Hammersley-Halton sequence: Let P = {p 1, p 2,..., } be the set of prime numbers in increasing order, i.e. p 1 = 2, p 2 = 3, p 3 = 5, p 4 = 7,.... Define quadrature points x, x 1,... by x n = (ϕ p1 (n), ϕ p2 (n),..., ϕ ps (n)) for n =, 1, 2,....
31 Hammersley-Halton sequence Hammerlsey-Halton point set with 64 points.
32 Digital nets Choose prime number b and finite field Z b = {, 1,..., b 1} of order b. Choose C 1,..., C s Z m m b. Let n = n + n 1 b + + n m 1 b m 1 and set n = (n,..., n m 1 ) Z m b. Let y n,i = C i n for 1 i s, n < b m. For y n,i = (y n,i,1,..., y n,i,m ) Z m b let x n,i = y n,i,1 b + + y n,i,m b m. Set x n = (x n,1,..., x n,s ) for n < b m.
33 (t, m, s)-net property Let m, s 1 and b 2 be integers. A point set P = {x,..., x bm 1} is called a (t, m, s)-net in base b, if for all integers d 1,..., d s with d d s = m t the number of points in the elementary intervals s i=1 [ ai b d i, a ) i + 1 b d i where a i < b d i, is b t.
34 If P = {x,..., x bm 1} [, 1] s is a (t, m, s)-net, then local discrepancy function ( a1 P b d,..., a ) s = 1 b ds for all a i < b d i, di, d d s = m t.
35
36
37
38
39
40
41
42
43
44
45
46
47 Sobol point set The first 64 points of a Sobol sequence which form a (, 6, 2)-net in base 2.
48 Niederreiter-Xing point set The first 64 points of a Niederreiter-Xing sequence.
49 Niederreiter-Xing point set The first 124 points of a Niederreiter-Xing sequence.
50 Kronecker sequence Let α 1,..., α s R be linearly independent over Q. Let x n = ({nα 1 },..., {nα s }) for n =, 1, 2,.... For instance, one can choose α i = p i where p i is the ith prime number.
51 Kronecker sequence The first 64 points of a Kronecker sequence.
52 Discrepancy bounds It is known that for all the constructions above one has (log N) c(s) L p (P) C s, for all N 1, 1 p, N where C s > and c(s) s. Lower bound: For all point sets P consisting of N points we have L p (P) C s (log N) (s 1)/2 N for all N 1, 1 < p. In comparison: random point set ( ) log log N E(L 2 (P)) = O. N
53 For 1 < p <, the exact order of convergence is known to be of order (log N) (s 1)/2. N Explicit constructions of such points were achieved by Chen & Skriganov for p = 2 and Skriganov for 1 < p <. Great open problem: What is the correct order of convergence of min P [,1] s P =N L (P)?
54 Randomized quasi-monte Carlo Introduce random element to deterministic point set. Has several advantages: yields an unbiased estimator; there is a statistical error estimate; better rate of convergence of the random-case error compared to worst-case error; performs similar to Monte Carlo for L 2 functions but has better rate of convergence for smooth functions;
55 Randomly shifted lattice rules Lattice rules can be randomized using a random shift: Lattice point set: {ng/n}, n =, 1,..., N 1. Shifted lattice rule: choose σ [, 1] s uniformly distributed; then the shifted lattice point set is given by {ng/n + σ}, n =, 1,..., N 1.
56 Lattice point set
57 Shifted lattice point set
58 Owen s scrambling Let Z b = {, 1,..., b 1} and let x = x 1 b + x 2 b 2 +, x 1, x 2,... Z b. Randomly choose permutations σ, σ x1, σ x1,x 2,... : Z b Z b. Let y 1 = σ(x 1 ) y 2 = σ x1 (x 2 ) y 3 = σ x1,x 2 (x 3 ) Set y = y 1 b + y 2 b 2 +.
59 Scrambled Sobol point set The first 124 points of a Sobol sequence (left) and a scrambled Sobol sequence (right).
60 Scrambled net variance Let Then e(r) = f (x) dx 1 [,1] N s E(e(R)) = b m 1 n= f (y n ). ( ) (log N) s Var(e(R)) = O N 2α+1 for integrands with smoothness α 1.
61 Dependence on the dimension Although the convergence rate is better for quasi-monte Carlo, there is a stronger dependence on the dimension: Monte Carlo: error = O(N 1/2 ); Quasi-Monte Carlo: error = O ( (log N) s 1 N ) ; Notice that g(n) := N 1 (log N) s 1 is an increasing function for N e s 1. So if s is large, say s 3, then N 1 (log N) 29 increases for N 1 12.
62 Intractable For integration error in H with reproducing kernel s K (x, y) = min{1 x i, 1 y i } i=1 it is known that e 2 (H, P) ( 1 N ( ) s ) 8 e 2 (H, ). 9 Error can only decrease if N > constant ( ) s 9. 8
63 Weighted function spaces (Sloan and Woźniakowski) Study weighted function spaces: Introduce γ 1, γ 2,..., > and define K (x, y) = Then if s (1 + γ i min{1 x i, 1 y i }). i=1 γ i < we have e(h s,γ, P) CN δ, where C > is independent of the dimension s and δ < 1. i=1
64 Tractability Minimal error over all methods using N points: Inverse of the error: for ε > ; Strong tractability: e N(H) = inf e(h, P); P: P =N N (s, ε) = min{n N : e N(H) εe (H)} for some constant C > ; N (s, ε) Cε β Sloan and Woźniakowski: For the function space considered above, we get strong tractability if γ i <. i=1
65 Tractability of star-discrepancy Recall the local discrepancy function P (t) = 1 N 1 1 [,t) (x n ) t 1 t s, where P = {x,..., x N 1 }. N n= The star-discrepancy is given by D N(P) = sup P (t). t [,1] s Result by Heinrich, Novak, Woźniakowski, Wasilkowski: s Minimal star discrepancy D (s, N) = inf P D N(P) C N for all s, N N. This implies that N (s, ε) Csε 2.
66 Extensions, open problems and future research Quasi-Monte Carlo methods which achieve convergence rates of order N α (log N) αs, α > 1 for sufficiently smooth integrands; Construction of point sets whose star-discrepancy is tractable; Completely uniformly distributed point sets to speed up convergence in Markov chain Monte Carlo algorithms; Infinite dimensional integration: Integrands have infinitely many variables; Connections to codes, orthogonal arrays and experimental designs; Quasi-Monte Carlo for function approximation; Choosing the weights in applications; Quasi-Monte Carlo on the sphere;
67 Book
68 Thank You! Special thanks to Dirk Nuyens for creating many of the pictures in this talk; Friedrich Pillichshammer for letting me use a picture of his daughters; Fred Hickernell, Peter Kritzer, Art Owen, and Friedrich Pillichshammer for helpful comments on the slides; All colleagues who worked on quasi-monte Carlo methods over the years.
New Multilevel Algorithms Based on Quasi-Monte Carlo Point Sets
New Multilevel Based on Quasi-Monte Carlo Point Sets Michael Gnewuch Institut für Informatik Christian-Albrechts-Universität Kiel 1 February 13, 2012 Based on Joint Work with Jan Baldeaux (UTS Sydney)
More informationLow Discrepancy Sequences in High Dimensions: How Well Are Their Projections Distributed?
Low Discrepancy Sequences in High Dimensions: How Well Are Their Projections Distributed? Xiaoqun Wang 1,2 and Ian H. Sloan 2 1 Department of Mathematical Sciences, Tsinghua University, Beijing 100084,
More informationDigital lattice rules
Digital lattice rules Multivariate integration and discrepancy estimates A thesis presented to The University of New South Wales in fulfilment of the thesis requirement for the degree of Doctor of Philosophy
More informationAPPLIED MATHEMATICS REPORT AMR04/2 DIAPHONY, DISCREPANCY, SPECTRAL TEST AND WORST-CASE ERROR. J. Dick and F. Pillichshammer
APPLIED MATHEMATICS REPORT AMR4/2 DIAPHONY, DISCREPANCY, SPECTRAL TEST AND WORST-CASE ERROR J. Dick and F. Pillichshammer January, 24 Diaphony, discrepancy, spectral test and worst-case error Josef Dick
More informationIan Sloan and Lattice Rules
www.oeaw.ac.at Ian Sloan and Lattice Rules P. Kritzer, H. Niederreiter, F. Pillichshammer RICAM-Report 2016-34 www.ricam.oeaw.ac.at Ian Sloan and Lattice Rules Peter Kritzer, Harald Niederreiter, Friedrich
More informationSTRONG TRACTABILITY OF MULTIVARIATE INTEGRATION USING QUASI MONTE CARLO ALGORITHMS
MATHEMATICS OF COMPUTATION Volume 72, Number 242, Pages 823 838 S 0025-5718(02)01440-0 Article electronically published on June 25, 2002 STRONG TRACTABILITY OF MULTIVARIATE INTEGRATION USING QUASI MONTE
More informationApplication of QMC methods to PDEs with random coefficients
Application of QMC methods to PDEs with random coefficients a survey of analysis and implementation Frances Kuo f.kuo@unsw.edu.au University of New South Wales, Sydney, Australia joint work with Ivan Graham
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 informationOptimal Randomized Algorithms for Integration on Function Spaces with underlying ANOVA decomposition
Optimal Randomized on Function Spaces with underlying ANOVA decomposition Michael Gnewuch 1 University of Kaiserslautern, Germany October 16, 2013 Based on Joint Work with Jan Baldeaux (UTS Sydney) & Josef
More informationOn the Behavior of the Weighted Star Discrepancy Bounds for Shifted Lattice Rules
On the Behavior of the Weighted Star Discrepancy Bounds for Shifted Lattice Rules Vasile Sinescu and Pierre L Ecuyer Abstract We examine the question of constructing shifted lattice rules of rank one with
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 informationHot new directions for QMC research in step with applications
Hot new directions for QMC research in step with applications Frances Kuo f.kuo@unsw.edu.au University of New South Wales, Sydney, Australia Frances Kuo @ UNSW Australia p.1 Outline Motivating example
More informationQuasi-Monte Carlo integration over the Euclidean space and applications
Quasi-Monte Carlo integration over the Euclidean space and applications f.kuo@unsw.edu.au University of New South Wales, Sydney, Australia joint work with James Nichols (UNSW) Journal of Complexity 30
More informationAPPLIED MATHEMATICS REPORT AMR04/16 FINITE-ORDER WEIGHTS IMPLY TRACTABILITY OF MULTIVARIATE INTEGRATION. I.H. Sloan, X. Wang and H.
APPLIED MATHEMATICS REPORT AMR04/16 FINITE-ORDER WEIGHTS IMPLY TRACTABILITY OF MULTIVARIATE INTEGRATION I.H. Sloan, X. Wang and H. Wozniakowski Published in Journal of Complexity, Volume 20, Number 1,
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 informationOn the existence of higher order polynomial. lattices with large figure of merit ϱ α.
On the existence of higher order polynomial lattices with large figure of merit Josef Dick a, Peter Kritzer b, Friedrich Pillichshammer c, Wolfgang Ch. Schmid b, a School of Mathematics, University of
More informationLattice rules and sequences for non periodic smooth integrands
Lattice rules and sequences for non periodic smooth integrands dirk.nuyens@cs.kuleuven.be Department of Computer Science KU Leuven, Belgium Joint work with Josef Dick (UNSW) and Friedrich Pillichshammer
More informationExponential Convergence and Tractability of Multivariate Integration for Korobov Spaces
Exponential Convergence and Tractability of Multivariate Integration for Korobov Spaces Josef Dick, Gerhard Larcher, Friedrich Pillichshammer and Henryk Woźniakowski March 12, 2010 Abstract In this paper
More informationImproved Discrepancy Bounds for Hybrid Sequences. Harald Niederreiter. RICAM Linz and University of Salzburg
Improved Discrepancy Bounds for Hybrid Sequences Harald Niederreiter RICAM Linz and University of Salzburg MC vs. QMC methods Hybrid sequences The basic sequences Deterministic discrepancy bounds The proof
More informationQuasi-Monte Carlo Methods for Applications in Statistics
Quasi-Monte Carlo Methods for Applications in Statistics Weights for QMC in Statistics Vasile Sinescu (UNSW) Weights for QMC in Statistics MCQMC February 2012 1 / 24 Quasi-Monte Carlo Methods for Applications
More informationDiscrepancy Bounds for Nets and Sequences
Discrepancy Bounds for Nets and Sequences Peter Kritzer Johann Radon Institute for Computational and Applied Mathematics (RICAM) Austrian Academy of Sciences Linz, Austria Joint work with Henri Faure (Marseille)
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 informationDiscrepancy Bounds for Deterministic Acceptance-Rejection Samplers
Discrepancy Bounds for Deterministic Acceptance-Rejection Samplers Houying Zhu Athesisinfulfilmentoftherequirementsforthedegreeof Doctor of Philosophy School of Mathematics and Statistics Faculty of Science
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 informationUniform Distribution and Quasi-Monte Carlo Methods
Workshop on Uniform Distribution and Quasi-Monte Carlo Methods October 14-18, 2013 as part of the Radon Special Semester 2013 on Applications of Algebra and Number Theory Metric number theory, lacunary
More informationA NEW UPPER BOUND ON THE STAR DISCREPANCY OF (0,1)-SEQUENCES. Peter Kritzer 1. Abstract
INTEGERS: ELECTRONIC JOURNAL OF COMBINATORIAL NUMBER THEORY 5(3 (2005, #A11 A NEW UPPER BOUND ON THE STAR DISCREPANCY OF (0,1-SEQUENCES Peter Kritzer 1 Department of Mathematics, University of Salzburg,
More informationWhat s new in high-dimensional integration? designing for applications
What s new in high-dimensional integration? designing for applications Ian H. Sloan i.sloan@unsw.edu.au The University of New South Wales UNSW Australia ANZIAM NSW/ACT, November, 2015 The theme High dimensional
More informationLifting the Curse of Dimensionality
Lifting the Curse of Dimensionality Frances Y. Kuo and Ian H. Sloan Introduction Richard Bellman [1] coined the phrase the curse of dimensionality to describe the extraordinarily rapid growth in the difficulty
More informationIntegration of permutation-invariant. functions
permutation-invariant Markus Weimar Philipps-University Marburg Joint work with Dirk Nuyens and Gowri Suryanarayana (KU Leuven, Belgium) MCQMC2014, Leuven April 06-11, 2014 un Outline Permutation-invariant
More informationStochastic optimization, multivariate numerical integration and Quasi-Monte Carlo methods. W. Römisch
Stochastic optimization, multivariate numerical integration and Quasi-Monte Carlo methods W. Römisch Humboldt-University Berlin Institute of Mathematics www.math.hu-berlin.de/~romisch Chemnitzer Mathematisches
More informationRandomized Quasi-Monte Carlo Simulation of Markov Chains with an Ordered State Space
Randomized Quasi-Monte Carlo Simulation of Markov Chains with an Ordered State Space Pierre L Ecuyer 1, Christian Lécot 2, and Bruno Tuffin 3 1 Département d informatique et de recherche opérationnelle,
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 informationEffective dimension of some weighted pre-sobolev spaces with dominating mixed partial derivatives
Effective dimension of some weighted pre-sobolev spaces with dominating mixed partial derivatives Art B. Owen Stanford University August 2018 Abstract This paper considers two notions of effective dimension
More informationApproximation of Functions Using Digital Nets
Approximation of Functions Using Digital Nets Josef Dick 1, Peter Kritzer 2, and Frances Y. Kuo 3 1 UNSW Asia, 1 Kay Siang Road, Singapore 248922. E-mail: j.dick@unswasia.edu.sg 2 Fachbereich Mathematik,
More informationConstruction algorithms for plane nets in base b
Construction algorithms for plane nets in base b Gunther Leobacher, Friedrich Pillichshammer and Thomas Schell arxiv:1506.03201v1 [math.nt] 10 Jun 2015 Abstract The class of (0, m, s)-nets in base b has
More informationOn the inverse of the discrepancy for infinite dimensional infinite sequences
On the inverse of the discrepancy for infinite dimensional infinite sequences Christoph Aistleitner Abstract In 2001 Heinrich, ovak, Wasilkowski and Woźniakowski proved the upper bound s,ε) c abs sε 2
More informationVariance and discrepancy with alternative scramblings
Variance and discrepancy with alternative scramblings ART B. OWEN Stanford University This paper analyzes some schemes for reducing the computational burden of digital scrambling. Some such schemes have
More informationMonte Carlo Method for Numerical Integration based on Sobol s Sequences
Monte Carlo Method for Numerical Integration based on Sobol s Sequences Ivan Dimov and Rayna Georgieva Institute of Information and Communication Technologies, Bulgarian Academy of Sciences, 1113 Sofia,
More informationSome Results on the Complexity of Numerical Integration
Some Results on the Complexity of Numerical Integration Erich Novak Abstract We present some results on the complexity of numerical integration. We start with the seminal paper of Bakhvalov (1959) and
More informationA Randomized Quasi-Monte Carlo Simulation Method for Markov Chains
A Randomized Quasi-Monte Carlo Simulation Method for Markov Chains Pierre L Ecuyer GERAD and Département d Informatique et de Recherche Opérationnelle Université de Montréal, C.P. 6128, Succ. Centre-Ville,
More informationApproximation of High-Dimensional Numerical Problems Algorithms, Analysis and Applications
Approximation of High-Dimensional Numerical Problems Algorithms, Analysis and Applications Christiane Lemieux (University of Waterloo), Ian Sloan (University of New South Wales), Henryk Woźniakowski (University
More informationA Belgian view on lattice rules
A Belgian view on lattice rules Ronald Cools Dept. of Computer Science, KU Leuven Linz, Austria, October 14 18, 2013 Atomium Brussels built in 1958 height 103m figure = 2e coin 5 10 6 in circulation Body
More informationLow-discrepancy point sets lifted to the unit sphere
Low-discrepancy point sets lifted to the unit sphere Johann S. Brauchart School of Mathematics and Statistics, UNSW MCQMC 2012 (UNSW) February 17 based on joint work with Ed Saff (Vanderbilt) and Josef
More informationFOCM Workshop B5 Information Based Complexity
FOCM 2014 - Workshop B5 Information Based Complexity B5 - December 15, 14:30 15:00 The ANOVA decomposition of a non-smooth function of an infinite number of variables Ian Sloan University of New South
More informationarxiv: v2 [math.na] 18 May 2011
Efficient calculation of the worst-case error and (fast component-by-component construction of higher order polynomial lattice rules arxiv:1105.2599v2 [math.na] 18 May 2011 Jan Baldeaux Josef Dick Gunther
More informationConstructions of digital nets using global function fields
ACTA ARITHMETICA 105.3 (2002) Constructions of digital nets using global function fields by Harald Niederreiter (Singapore) and Ferruh Özbudak (Ankara) 1. Introduction. The theory of (t, m, s)-nets and
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 informationQuadrature using sparse grids on products of spheres
Quadrature using sparse grids on products of spheres Paul Leopardi Mathematical Sciences Institute, Australian National University. For presentation at 3rd Workshop on High-Dimensional Approximation UNSW,
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 informationDyadic diaphony of digital sequences
Dyadic diaphony of digital sequences Friedrich Pillichshammer Abstract The dyadic diaphony is a quantitative measure for the irregularity of distribution of a sequence in the unit cube In this paper we
More informationANOVA decomposition of convex piecewise linear functions
ANOVA decomposition of convex piecewise linear functions W. Römisch Abstract Piecewise linear convex functions arise as integrands in stochastic programs. They are Lipschitz continuous on their domain,
More information2 FRED J. HICKERNELL the sample mean of the y (i) : (2) ^ 1 N The mean square error of this estimate may be written as a sum of two parts, a bias term
GOODNESS OF FIT STATISTICS, DISCREPANCIES AND ROBUST DESIGNS FRED J. HICKERNELL Abstract. The Cramer{Von Mises goodness-of-t statistic, also known as the L 2 -star discrepancy, is the optimality criterion
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 informationA Randomized Quasi-Monte Carlo Simulation Method for Markov Chains
A Randomized Quasi-Monte Carlo Simulation Method for Markov Chains Pierre L Ecuyer GERAD and Département d Informatique et de Recherche Opérationnelle Université de Montréal, C.P. 6128, Succ. Centre-Ville,
More informationQUASI-MONTE CARLO VIA LINEAR SHIFT-REGISTER SEQUENCES. Pierre L Ecuyer Christiane Lemieux
Proceedings of the 1999 Winter Simulation Conference P. A. Farrington, H. B. Nembhard, D. T. Sturrock, and G. W. Evans, eds. QUASI-MONTE CARLO VIA LINEAR SHIFT-REGISTER SEQUENCES Pierre L Ecuyer Christiane
More informationError Analysis for Quasi-Monte Carlo Methods
Error Analysis for Quasi-Monte Carlo Methods Fred J. Hickernell Department of Applied Mathematics, Illinois Institute of Technology hickernell@iit.edu mypages.iit.edu/~hickernell Thanks to the Guaranteed
More informationA Coding Theoretic Approach to Building Nets with Well-Equidistributed Projections
A Coding Theoretic Approach to Building Nets with Well-Equidistributed Projections Yves Edel 1 and Pierre L Ecuyer 2 1 Mathematisches Institut der Universität Im Neuenheimer Feld 288, 69120 Heidelberg,
More informationReliable Error Estimation for Quasi-Monte Carlo Methods
Reliable Error Estimation for Quasi-Monte Carlo Methods Department of Applied Mathematics, Joint work with Fred J. Hickernell July 8, 2014 Outline 1 Motivation Reasons Problem 2 New approach Set-up of
More informationConstruction Algorithms for Higher Order Polynomial Lattice Rules
Construction Algorithms for Higher Order Polynomial Lattice Rules Jan Baldeaux, Josef Dick, Julia Greslehner and Friedrich Pillichshammer Dedicated to Gerhard Larcher on the occasion of his 50th birthday
More informationReport on Dagstuhl-Seminar 00391: Algorithms and Complexity for Continuous Problems. September 25-29, 2000
Report on Dagstuhl-Seminar 00391: Algorithms and Complexity for Continuous Problems September 25-29, 2000 The seminar was attended by 36 scientists from 12 countries. It was devoted to the complexity of
More informationQUASI-MONTE CARLO METHODS FOR SIMULATION. Pierre L Ecuyer
Proceedings of the 2003 Winter Simulation Conference S Chick, P J Sánchez, D Ferrin, and D J Morrice, eds QUASI-MONTE CARLO METHODS FOR SIMULATION Pierre L Ecuyer Département d Informatique et de Recherche
More informationAdvanced Algorithms for Multidimensional Sensitivity Studies of Large-scale Air Pollution Models Based on Sobol Sequences
Advanced Algorithms for Multidimensional Sensitivity Studies of Large-scale Air Pollution Models Based on Sobol Sequences I. Dimov a,, R. Georgieva a, Tz. Ostromsky a, Z. Zlatev b a Department of Parallel
More informationFinding Optimal Volume Subintervals with k Points and Calculating the Star Discrepancy are NP-Hard Problems
Finding Optimal Volume Subintervals with k Points and Calculating the Star Discrepancy are NP-Hard Problems Michael Gnewuch Department of Computer Science, Kiel University, Christian-Albrechts-Platz 4,
More informationA PIECEWISE CONSTANT ALGORITHM FOR WEIGHTED L 1 APPROXIMATION OVER BOUNDED OR UNBOUNDED REGIONS IN R s
A PIECEWISE CONSTANT ALGORITHM FOR WEIGHTED L 1 APPROXIMATION OVER BONDED OR NBONDED REGIONS IN R s FRED J. HICKERNELL, IAN H. SLOAN, AND GRZEGORZ W. WASILKOWSKI Abstract. sing Smolyak s construction [5],
More informationRandom Number Generation: RNG: A Practitioner s Overview
Random Number Generation: A Practitioner s Overview Department of Computer Science Department of Mathematics Department of Scientific Computing Florida State University, Tallahassee, FL 32306 USA E-mail:
More informationSimulation Estimation of Mixed Discrete Choice Models Using Randomized and Scrambled Halton Sequences
Simulation Estimation of Mixed Discrete Choice Models Using Randomized and Scrambled Halton Sequences Chandra R. Bhat Dept of Civil Engineering, ECJ 6.8, Univ. of Texas at Austin, Austin, TX, 78712 Phone:
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 informationQMC tutorial 1. Quasi-Monte Carlo. Art B. Owen Stanford University. A tutorial reflecting personal views on what is important in QMC.
QMC tutorial 1 Quasi-Monte Carlo Art B. Owen Stanford University A tutorial reflecting personal views on what is important in QMC. MCQMC 2016 will be at Stanford, August 14-19 mcqmc2016.stanford.edu QMC
More informationApproximation numbers of Sobolev embeddings - Sharp constants and tractability
Approximation numbers of Sobolev embeddings - Sharp constants and tractability Thomas Kühn Universität Leipzig, Germany Workshop Uniform Distribution Theory and Applications Oberwolfach, 29 September -
More informationarxiv: v2 [math.na] 15 Aug 2018
Noname manuscript No. (will be inserted by the editor) Lattice rules in non-periodic subspaces of Sobolev spaces Takashi Goda Kosuke Suzuki Takehito Yoshiki arxiv:72.2572v2 [math.na] 5 Aug 28 Received:
More informationBMO and exponential Orlicz space estimates of the discrepancy function in arbitrary dimension
BMO and exponential Orlicz space estimates of the discrepancy function in arbitrary dimension arxiv:1411.5794v2 [math.fa] 9 Jul 2015 Dmitriy Bilyk a, Lev Markhasin b a School of Mathematics, University
More informationON TRACTABILITY OF WEIGHTED INTEGRATION OVER BOUNDED AND UNBOUNDED REGIONS IN R s
MATHEMATICS OF COMPTATION Volume 73, Number 248, Pages 1885 1901 S 0025-5718(04)01624-2 Article electronically published on January 5, 2004 ON TRACTABILITY OF WEIGHTED INTEGRATION OVER BONDED AND NBONDED
More informationThe square root rule for adaptive importance sampling
The square root rule for adaptive importance sampling Art B. Owen Stanford University Yi Zhou January 2019 Abstract In adaptive importance sampling, and other contexts, we have unbiased and uncorrelated
More informationQMC designs and covering of spheres by spherical caps
QMC designs and covering of spheres by spherical caps Ian H. Sloan University of New South Wales, Sydney, Australia Joint work with Johann Brauchart, Josef Dick, Ed Saff (Vanderbilt), Rob Womersley and
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 informationConstruction algorithms for good extensible lattice rules
Construction algorithms for good extensible lattice rules Harald Niederreiter and Friedrich Pillichshammer Dedicated to Ian H. Sloan on the occasion of his 70th birthday Abstract Extensible olynomial lattice
More informationL 2 Discrepancy of Two-Dimensional Digitally Shifted Hammersley Point Sets in Base b
L Discrepancy of Two-Dimensional Digitally Shifted Hammersley Point Sets in Base Henri Faure and Friedrich Pillichshammer Astract We give an exact formula for the L discrepancy of two-dimensional digitally
More informationScenario generation in stochastic programming
Scenario generation in stochastic programming Werner Römisch Humboldt-University Berlin, Institute of Mathematics, 10099 Berlin, Germany, romisch@math.hu-berlin.de Abstract Stability-based methods for
More informationWorkshop on the Occasion of Harald Niederreiter's 70th Birthday: Applications of Algebra and Number Theory
Workshop on the Occasion of Harald Niederreiter's 70th Birthday: Applications of Algebra and Number Theory June 23-27, 2014 Harald Niederreiter in Marseille Monday Tuesday Wednesday Thursday Friday 08:15
More informationQuadrature using sparse grids on products of spheres
Quadrature using sparse grids on products of spheres Paul Leopardi Mathematical Sciences Institute, Australian National University. For presentation at ANZIAM NSW/ACT Annual Meeting, Batemans Bay. Joint
More informationGeneralized Halton Sequences in 2008: A Comparative Study
Generalized alton Sequences in 2008: A Comparative Study ENRI FAURE Institut Mathématiques de Luminy and Université Paul Cézanne and CRISTIANE LEMIEUX University of Waterloo alton sequences have always
More informationQuasi-Random Simulation of Discrete Choice Models
Quasi-Random Simulation of Discrete Choice Models by Zsolt Sándor and Kenneth Train Erasmus University Rotterdam and University of California, Berkeley June 12, 2002 Abstract We describe the properties
More informationScenario generation Contents
Scenario generation W. Römisch Humboldt-University Berlin Department of Mathematics www.math.hu-berlin.de/~romisch Page 1 of 39 Tutorial, ICSP12, Halifax, August 14, 2010 (1) Aproximation issues (2) Generation
More informationOn lower bounds for integration of multivariate permutation-invariant functions
On lower bounds for of multivariate permutation-invariant functions Markus Weimar Philipps-University Marburg Oberwolfach October 2013 Research supported by Deutsche Forschungsgemeinschaft DFG (DA 360/19-1)
More informationRandomized Quasi-Monte Carlo for MCMC
Randomized Quasi-Monte Carlo for MCMC Radu Craiu 1 Christiane Lemieux 2 1 Department of Statistics, Toronto 2 Department of Statistics, Waterloo Third Workshop on Monte Carlo Methods Harvard, May 2007
More informationFAST QMC MATRIX-VECTOR MULTIPLICATION
FAST QMC MATRIX-VECTOR MULTIPLICATION JOSEF DICK, FRANCES Y. KUO, QUOC T. LE GIA, CHRISTOPH SCHWAB Abstract. Quasi-Monte Carlo (QMC) rules 1/N N 1 n=0 f(y na) can be used to approximate integrals of the
More informationIMPROVEMENT ON THE DISCREPANCY. 1. Introduction
t m Mathematical Publications DOI: 10.2478/tmmp-2014-0016 Tatra Mt. Math. Publ. 59 (2014 27 38 IMPROVEMENT ON THE DISCREPANCY OF (t es-sequences Shu Tezuka ABSTRACT. Recently a notion of (t es-sequences
More informationIntroduction to (randomized) quasi-monte Carlo
1 aft Introduction to (randomized) quasi-monte Carlo Dr Pierre L Ecuyer MCQMC Conference, Stanford University, August 2016 Program Monte Carlo, Quasi-Monte Carlo, Randomized quasi-monte Carlo QMC point
More informationarxiv: v2 [math.na] 15 Oct 2015
Optimal quasi-monte Carlo rules on order 2 digital nets for the numerical integration of multivariate periodic functions Aicke Hinrichs a, Lev Markhasin b, Jens Oettershagen c, Tino Ullrich c arxiv:1501.01800v2
More informationLow Discrepancy Toolbox Manual. Michael Baudin
Low Discrepancy Toolbox Manual Michael Baudin Version 0.1 May 2013 Abstract This document is an introduction to the Low Discrepancy module v0.5. Contents 1 Gallery 4 1.1 The Van Der Corput sequence.............................
More informationQuasi-Monte Carlo Sampling by Art B. Owen
Chapter 1 Quasi-Monte Carlo Sampling by Art B. Owen In Monte Carlo (MC) sampling the sample averages of random quantities are used to estimate the corresponding expectations. The justification is through
More informationApproximation of stochastic optimization problems and scenario generation
Approximation of stochastic optimization problems and scenario generation W. Römisch Humboldt-University Berlin Institute of Mathematics 10099 Berlin, Germany http://www.math.hu-berlin.de/~romisch Page
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 informationImproved Halton sequences and discrepancy bounds
Monte Carlo Methods Appl. Vol. No. ( ), pp. 1 18 DOI 10.1515 / MCMA.. c de Gruyter Improved Halton sequences and discrepancy bounds Henri Faure and Christiane Lemieux Abstract. For about fifteen years,
More informationFast evaluation of mixed derivatives and calculation of optimal weights for integration. Hernan Leovey
Fast evaluation of mixed derivatives and calculation of optimal weights for integration Humboldt Universität zu Berlin 02.14.2012 MCQMC2012 Tenth International Conference on Monte Carlo and Quasi Monte
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 informationAdapting quasi-monte Carlo methods to simulation problems in weighted Korobov spaces
Adapting quasi-monte Carlo methods to simulation problems in weighted Korobov spaces Christian Irrgeher joint work with G. Leobacher RICAM Special Semester Workshop 1 Uniform distribution and quasi-monte
More informationarxiv: v1 [math.na] 13 Jul 2017
Richardson extrapolation of polynomial lattice rules Josef Dick, Takashi Goda, Takehito Yoshiki arxiv:707.03989v [math.na] 3 Jul 207 July 4, 207 Abstract We study multivariate numerical integration of
More informationQUASI-MONTE CARLO METHODS IN NON-CUBICAL SPACES
QUASI-MONTE CARLO METHODS IN NON-CUBICAL SPACES A DISSERTATION SUBMITTED TO THE DEPARTMENT OF STATISTICS AND THE COMMITTEE ON GRADUATE STUDIES OF STANFORD UNIVERSITY IN PARTIAL FULFILLMENT OF THE REQUIREMENTS
More informationQuasi-Monte Carlo methods with applications in finance
Finance Stoch (2009) 13: 307 349 DOI 10.1007/s00780-009-0095-y Quasi-Monte Carlo methods with applications in finance Pierre L Ecuyer Received: 30 May 2008 / Accepted: 20 August 2008 / Published online:
More information