negative dependence stable polynomials etc in ml
play

Negative Dependence, Stable Polynomials etc in ML Part 2 SUVRIT - PowerPoint PPT Presentation

Mar 28, 2023 •461 likes •787 views

Negative Dependence, Stable Polynomials etc in ML Part 2 SUVRIT SRA & STEFANIE JEGELKA Laboratory for Information and Decision Systems Massachusetts Institute of Technology Neural information Processing Systems, 2018 ml.mit.edu

  1. Negative Dependence, 
 Stable Polynomials etc in ML Part 2 SUVRIT SRA & STEFANIE JEGELKA Laboratory for Information and Decision Systems Massachusetts Institute of Technology Neural information Processing Systems, 2018 ml.mit.edu

  2. Outline Introduction 
 Prominent example: Determinantal Point Processes 1 Stronger notions of negative dependence Intro & Implications: Sampling Theory Approximating partition functions Learning a DPP (and some variants) 2 Applications Theory & Applications Recommender systems, Nyström method, optimal design, regression, neural net pruning, negative mining, anomaly detection, etc. Perspectives and wrap-up Negative dependence, stable polynomials etc. in ML - part 1 Stefanie Jegelka (stefje@mit.edu)

  3. Theory Partition functions Learning DPPs Negative dependence, stable polynomials etc. in ML - part 2 Suvrit Sra (suvrit@mit.edu)

  4. <latexit sha1_base64="qwEaZ3oTpAJyVs2s06lKeQV+gtY=">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</latexit> <latexit sha1_base64="qwEaZ3oTpAJyVs2s06lKeQV+gtY=">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</latexit> <latexit sha1_base64="qwEaZ3oTpAJyVs2s06lKeQV+gtY=">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</latexit> <latexit sha1_base64="qwEaZ3oTpAJyVs2s06lKeQV+gtY=">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</latexit> Computing Partition functions Aim: Estimate Z µ , i.e., normalization const / partition function Pr( S ) = 1 µ ( S ) Z µ Typically intractable and often even hard to approximate (exponential number of terms to sum over, or evaluation of high-dimensional integrals / volumes) but… 4 Negative dependence, stable polynomials etc. in ML - part 2 Suvrit Sra (suvrit@mit.edu)

Download Presentation
Download Policy: The content available on the website is offered to you 'AS IS' for your personal information and use only. It cannot be commercialized, licensed, or distributed on other websites without prior consent from the author. To download a presentation, simply click this link. If you encounter any difficulties during the download process, it's possible that the publisher has removed the file from their server.

Recommend

Negative Dependence, Stable Polynomials etc in ML Part 1 STEFANIE JEGELKA &amp; SUVRIT SRA

Negative Dependence, Stable Polynomials etc in ML Part 1 STEFANIE JEGELKA &amp; SUVRIT SRA

Negative Dependence, Stable Polynomials etc in ML Part 1 STEFANIE JEGELKA &amp; SUVRIT SRA Dept of EECS &amp; CSAIL Massachusetts Institute of Technology Neural information Processing Systems, 2018 ml.mit.edu Negative Dependence

350 views • 34 slides
The symmetry and Schur expansion of dual stable Grothendieck polynomials Pavel Galashin MIT

The symmetry and Schur expansion of dual stable Grothendieck polynomials Pavel Galashin MIT

The symmetry and Schur expansion of dual stable Grothendieck polynomials Pavel Galashin MIT October 7, 2015 Joint work with Gaku Liu and Darij Grinberg Pavel Galashin (MIT) Dual stable Grothendieck polynomials October 7, 2015 1 / 25 Part

1.12k views • 95 slides
Refined dual stable Grothendieck polynomials and generalized Bender-Knuth involutions Gaku Liu

Refined dual stable Grothendieck polynomials and generalized Bender-Knuth involutions Gaku Liu

What is a dual stable Grothendieck polynomial? Generalized BK involutions Refined dual stable Grothendieck polynomials Refined dual stable Grothendieck polynomials and generalized Bender-Knuth involutions Gaku Liu Joint work with Pavel

837 views • 69 slides
Refined dual stable Grothendieck polynomials and generalized Bender-Knuth involutions Pavel

Refined dual stable Grothendieck polynomials and generalized Bender-Knuth involutions Pavel

DMTCS proc. BC , 2016, 527538 FPSAC 2016 Vancouver, Canada Refined dual stable Grothendieck polynomials and generalized Bender-Knuth involutions Pavel Galashin, Darij Grinberg, and Gaku Liu Massachusetts Institute of Technology, USA

275 views • 12 slides
Hochberg Multiple Test Procedure Under Negative Dependence Ajit C. Tamhane Northwestern

Hochberg Multiple Test Procedure Under Negative Dependence Ajit C. Tamhane Northwestern

Preliminaries Conservative Simes Test Multivariate Uniform Distribution Models Error Rate Control for n = 2 Error Rate Control Hochberg Multiple Test Procedure Under Negative Dependence Ajit C. Tamhane Northwestern University Joint work with

491 views • 27 slides
Control-dependence Analysis 2 Control-dependence Analysis 1. Introduction (motivation, overview)

Control-dependence Analysis 2 Control-dependence Analysis 1. Introduction (motivation, overview)

Class 4 Basic Analyses (4) Assign (see Schedule for links) Readings Control/program-dependence analysis Static single assignment and control dependence Problem Set 2: due 9/1/09 1 Control-dependence Analysis 2

254 views • 12 slides
Linear dependence and independence Linear dependence 1 Definition (linear (in)dependence) Let {

Linear dependence and independence Linear dependence 1 Definition (linear (in)dependence) Let {

L INEAR A LGEBRA Berkant Ustao glu CRYPTOLOUNGE . NET Linear dependence and independence Linear dependence 1 Definition (linear (in)dependence) Let { v 1 , v 2 , . . . , v k } be a set of vectors. If v k = a 1 v 1 + a 2

382 views • 25 slides
Measuring Dependence and Conditional Dependence with Kernels Kenji Fukumizu The Institute of

Measuring Dependence and Conditional Dependence with Kernels Kenji Fukumizu The Institute of

Measuring Dependence and Conditional Dependence with Kernels Kenji Fukumizu The Institute of Statistical Mathematics, Japan June 25, 2014. ICML 2014, Causality Workshop 1 2 Introduction Dependence Measures Dependence measures and

950 views • 32 slides
More Zeroes of Polynomials In this lecture we look more carefully at zeroes of polynomials.

More Zeroes of Polynomials In this lecture we look more carefully at zeroes of polynomials.

More Zeroes of Polynomials In this lecture we look more carefully at zeroes of polynomials. (Recall: a zero of a polynomial is sometimes called a root.) Our goal in the next few presentations is to set up a strategy for Elementary

331 views • 9 slides
28 4x 81y 4 z rt 3 2 , 4 5 6 7 2 17x mn 3 We usually write the variables in exponential

28 4x 81y 4 z rt 3 2 , 4 5 6 7 2 17x mn 3 We usually write the variables in exponential

Slide 1 / 216 Slide 2 / 216 Algebra I Polynomials 2015-11-02 www.njctl.org Slide 3 / 216 Click on the topic to go to that section Table of Contents Definitions of Monomials, Polynomials and Degrees Adding and Subtracting Polynomials

964 views • 72 slides
ASSA ABLOY Q4 Report 2019 Public Stable growth and strong EBIT in Q4 2019 Organic sales

ASSA ABLOY Q4 Report 2019 Public Stable growth and strong EBIT in Q4 2019 Organic sales

ASSA ABLOY Q4 Report 2019 Public Stable growth and strong EBIT in Q4 2019 Organic sales development Strong growth in Americas Good growth in Global Technologies Stable growth in EMEA and Entrance Systems Negative growth in

551 views • 34 slides
Classification of Combinatorial Polynomials (in particular, Ehrhart Polynomials of Zonotopes)

Classification of Combinatorial Polynomials (in particular, Ehrhart Polynomials of Zonotopes)

Classification of Combinatorial Polynomials (in particular, Ehrhart Polynomials of Zonotopes) Matthias Beck San Francisco State University Katharina Jochemko Kungliga Tekniska H ogskolan Emily McCullough University of San Francisco

244 views • 20 slides
Unimodality of q -Eulerian polynomials and q , p -Eulerian polynomials Michelle Wachs University

Unimodality of q -Eulerian polynomials and q , p -Eulerian polynomials Michelle Wachs University

Unimodality of q -Eulerian polynomials and q , p -Eulerian polynomials Michelle Wachs University of Miami Joint work with John Shareshian and with Anthony Henderson Eulerian numbers and Eulerian polynomials Eulerian numbers: a n , j := |{

710 views • 43 slides
Advances in Knot Polynomials 21 October 2016 Advances in Knot Polynomials 21 October 2016 1 /

Advances in Knot Polynomials 21 October 2016 Advances in Knot Polynomials 21 October 2016 1 /

Advances in Knot Polynomials 21 October 2016 Advances in Knot Polynomials 21 October 2016 1 / 49 ABSTRACT Review of achievements and problems in the theory of colored knot polynomials. Accent is on the current mystery around the differential

761 views • 49 slides
Robust Algorithms for Chebyshev Polynomials and Related Approximations Miroslav Vl cek

Robust Algorithms for Chebyshev Polynomials and Related Approximations Miroslav Vl cek

Introduction Chebyshev polynomials Symmetrical Zolotarev polynomials Overview Application in Filter Design Conclusion Robust Algorithms for Chebyshev Polynomials and Related Approximations Miroslav Vl cek vlcek@fd.cvut.cz Department of

484 views • 37 slides
More refined representations Control dependence graph Problem: control-flow edges in CFG

More refined representations Control dependence graph Problem: control-flow edges in CFG

More refined representations Control dependence graph Problem: control-flow edges in CFG overspecify evaluation Program dependence graph (PDG): order data dependence graph + control dependence graph (CDG) [Ferrante, Ottenstein, &amp; Warren,

260 views • 5 slides
B. Schwenker Universitt Gttingen For the test beam crew Small PXD9 @ DESY 2015 First

B. Schwenker Universitt Gttingen For the test beam crew Small PXD9 @ DESY 2015 First

Summary of TB results for the small PXD9 matrix B. Schwenker Universitt Gttingen For the test beam crew Small PXD9 @ DESY 2015 First Belle II type matrix in a test beam integrated into EUDET telescope PXD9 small Belle II type

549 views • 32 slides
FRACTIONAL UNDERDAMPED LANGEVIN DYNAMICS: Umut im ekli LTCI, Tlcom Paris, RETARGETING

FRACTIONAL UNDERDAMPED LANGEVIN DYNAMICS: Umut im ekli LTCI, Tlcom Paris, RETARGETING

FRACTIONAL UNDERDAMPED LANGEVIN DYNAMICS: Umut im ekli LTCI, Tlcom Paris, RETARGETING SGD WITH MOMENTUM UNDER Institut Polytechnique de Paris HEAVY-TAILED GRADIENT NOISE Umut im ekli*, Lingjiong Zhu*, Yee Whye Teh, Mert

402 views • 15 slides
Kakutani equivalence of unipotent flows Adam Kanigowski Montreal, 07.27.2018 (joint w. K.

Kakutani equivalence of unipotent flows Adam Kanigowski Montreal, 07.27.2018 (joint w. K.

Kakutani equivalence of unipotent flows Adam Kanigowski Montreal, 07.27.2018 (joint w. K. Vinhage and D. Wei) 1 / 11 General setting ( X , B , ) probability standard Borel space; ( T t ) : ( X , B , ) ( X , B , )

675 views • 40 slides
UT DA Power Grid Reduction by Sparse Convex Optimization Wei Ye 1 , Meng Li 1 , Kai Zhong 2 ,

UT DA Power Grid Reduction by Sparse Convex Optimization Wei Ye 1 , Meng Li 1 , Kai Zhong 2 ,

UT DA Power Grid Reduction by Sparse Convex Optimization Wei Ye 1 , Meng Li 1 , Kai Zhong 2 , Bei Yu 3 , David Z. Pan 1 1 ECE Department, University of Texas at Austin 2 ICES, University of Texas at Austin 3 CSE Department, Chinese University of

796 views • 22 slides
Inequivalent bundle representations for the Noncommutative Torus Chern numbers: from

Inequivalent bundle representations for the Noncommutative Torus Chern numbers: from

Inequivalent bundle representations for the Noncommutative Torus Chern numbers: from abstract to concrete Giuseppe De Nittis Mathematical Physics Sector of: SISSA International School for Advanced Studies, Trieste Noncommutative

662 views • 47 slides
Evaluation of the ARC Controller for the NASA IRTF Anthony Denault, Charles Lockhart, Eric

Evaluation of the ARC Controller for the NASA IRTF Anthony Denault, Charles Lockhart, Eric

Evaluation of the ARC Controller for the NASA IRTF Anthony Denault, Charles Lockhart, Eric Warmbier, &amp; John Rayner 2012.01.24 Our goal is to answer this question: Does the ARC Gen III array controller satisfy IRTF's needs? Noise

631 views • 39 slides
Introduction (1) Packet Loss Recovery for Streaming is growing Commercial streaming

Introduction (1) Packet Loss Recovery for Streaming is growing Commercial streaming

Introduction (1) Packet Loss Recovery for Streaming is growing Commercial streaming successful (ie Streaming Video RealPlayer and MediaPlayer) but proprietary and inflexible N. Feamster and H. Balakrishnan Use MPEG-4 since open

478 views • 6 slides
SCADA Security: Why is it so hard? Amol Sarwate, Director of Vulnerability Labs, Qualys Inc.

SCADA Security: Why is it so hard? Amol Sarwate, Director of Vulnerability Labs, Qualys Inc.

SCADA Security: Why is it so hard? Amol Sarwate, Director of Vulnerability Labs, Qualys Inc. June 22, 2012 HACK IN PARIS Agenda SCADA Basics Threats (where, why &amp; how) Challenges Recommendations and Proposals ScadaScan tool HACK

997 views • 65 slides

More recommend