risk minimum risk training reinforcement learning
play

Risk, Minimum Risk Training, Reinforcement Learning Graham Neubig - PowerPoint PPT Presentation

Sep 02, 2023 •312 likes •710 views

CS11-747 Neural Networks for NLP Risk, Minimum Risk Training, Reinforcement Learning Graham Neubig Site https://phontron.com/class/nn4nlp2020/ <latexit

  1. CS11-747 Neural Networks for NLP Risk, Minimum Risk Training, Reinforcement Learning Graham Neubig Site https://phontron.com/class/nn4nlp2020/

  2. <latexit sha1_base64="GeA/Os4/BK6Zz954iZvfPtPrQE=">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</latexit> <latexit sha1_base64="GeA/Os4/BK6Zz954iZvfPtPrQE=">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</latexit> <latexit sha1_base64="GeA/Os4/BK6Zz954iZvfPtPrQE=">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</latexit> Maximum Likelihood Training • Maximum the likelihood of predicting the next word in the reference given the previous words ` ( E | F ) = − log P ( E | F ) T X = − log P ( e t | F, e 1 , . . . , e t − 1 ) t =1 • Also called "teacher forcing"

  3. Problem 1: Exposure Bias • Teacher forcing assumes feeding correct previous input, but at test time we may make mistakes that propagate encoder I I I I classify classify classify classify classify I I I I I • Exposure bias: The model is not exposed to mistakes during training, and cannot deal with them at test

  4. Problem 2: Disregard to Evaluation Metrics • In the end, we want good outputs • Good translations can be measured with metrics, e.g. BLEU or METEOR • Some mistaken predictions hurt more than others, so we'd like to penalize them appropriately

  5. Error and Risk

  6. 
 
 <latexit sha1_base64="6ek90mJoNPTvCtomTW+aydQsu2s=">ACH3icbVBNSwMxEM3W7/pV9eglWAS9lF0RqgehKIrHClaFbinZ7LQNTXaXZFZalv0pXvwrXjyoiDf/jWkt4teDwJv3ZpjMCxIpDLru1OYmp6ZnZtfKC4uLa+sltbWr0ycag4NHstY3wTMgBQRNFCghJtEA1OBhOugfzLyr29BGxFHlzhMoKVYNxIdwRlaqV2q+j2G2WlOj6iPMCM6a5ig7yd+ShkCNbKaX3nq6C+EiE92Xym7FHYP+Jd6ElMkE9XbpzQ9jniqIkEtmTNzE2zZdSi4hLzopwYSxvusC01LI6bAtLxgTndtkpIO7G2L0I6Vr9PZEwZM1SB7VQMe+a3NxL/85opdg5amYiSFCHin4s6qaQY01FaNBQaOMqhJYxrYf9KeY9pxtFmWrQheL9P/ksae5XDinexX64dT9KYJ5tki+wQj1RJjZyTOmkQTu7IA3kiz8698+i8OK+frQVnMrNBfsB5/wAY9KMb</latexit> <latexit sha1_base64="6ek90mJoNPTvCtomTW+aydQsu2s=">ACH3icbVBNSwMxEM3W7/pV9eglWAS9lF0RqgehKIrHClaFbinZ7LQNTXaXZFZalv0pXvwrXjyoiDf/jWkt4teDwJv3ZpjMCxIpDLru1OYmp6ZnZtfKC4uLa+sltbWr0ycag4NHstY3wTMgBQRNFCghJtEA1OBhOugfzLyr29BGxFHlzhMoKVYNxIdwRlaqV2q+j2G2WlOj6iPMCM6a5ig7yd+ShkCNbKaX3nq6C+EiE92Xym7FHYP+Jd6ElMkE9XbpzQ9jniqIkEtmTNzE2zZdSi4hLzopwYSxvusC01LI6bAtLxgTndtkpIO7G2L0I6Vr9PZEwZM1SB7VQMe+a3NxL/85opdg5amYiSFCHin4s6qaQY01FaNBQaOMqhJYxrYf9KeY9pxtFmWrQheL9P/ksae5XDinexX64dT9KYJ5tki+wQj1RJjZyTOmkQTu7IA3kiz8698+i8OK+frQVnMrNBfsB5/wAY9KMb</latexit> <latexit sha1_base64="6ek90mJoNPTvCtomTW+aydQsu2s=">ACH3icbVBNSwMxEM3W7/pV9eglWAS9lF0RqgehKIrHClaFbinZ7LQNTXaXZFZalv0pXvwrXjyoiDf/jWkt4teDwJv3ZpjMCxIpDLru1OYmp6ZnZtfKC4uLa+sltbWr0ycag4NHstY3wTMgBQRNFCghJtEA1OBhOugfzLyr29BGxFHlzhMoKVYNxIdwRlaqV2q+j2G2WlOj6iPMCM6a5ig7yd+ShkCNbKaX3nq6C+EiE92Xym7FHYP+Jd6ElMkE9XbpzQ9jniqIkEtmTNzE2zZdSi4hLzopwYSxvusC01LI6bAtLxgTndtkpIO7G2L0I6Vr9PZEwZM1SB7VQMe+a3NxL/85opdg5amYiSFCHin4s6qaQY01FaNBQaOMqhJYxrYf9KeY9pxtFmWrQheL9P/ksae5XDinexX64dT9KYJ5tki+wQj1RJjZyTOmkQTu7IA3kiz8698+i8OK+frQVnMrNBfsB5/wAY9KMb</latexit> <latexit sha1_base64="KRxJjxRAFBSumCLgm+mSm7rf7k=">ACHicbVDLSgNBEJyNrxhfUY9eBoMQcOuBNSDECIBDx4iuEZIQpidJIhsw9mesWw5Ee8+CtePKh48SD4N04eB40WNBRV3XR3eZEUGm37y0rNzS8sLqWXMyura+sb2c2tGx3GioPLQxmqW49pkCIAFwVKuI0UMN+TUP65yO/dgdKizC4xkETZ91A9ERnKGRWtliA+EeE1AqVMN85YA2egyTynCfnlHncGKWLyvuL6+VzdkFewz6lzhTkiNTVFvZj0Y75LEPAXLJtK47doTNhCkUXMIw04g1RIz3WRfqhgbMB91Mxt8N6Z5R2rQTKlMB0rH6cyJhvtYD3zOdPsOenvVG4n9ePcbOSTMRQRQjBHyqBNLiEdRUXbQgFHOTCEcSXMrZT3mGIcTaAZE4Iz+/Jf4h4VTgvOVTFXKk/TSJMdskvyxCHpEQuSJW4hJMH8kReyKv1aD1b9b7pDVlTWe2yS9Yn9F6BW</latexit> <latexit sha1_base64="KRxJjxRAFBSumCLgm+mSm7rf7k=">ACHicbVDLSgNBEJyNrxhfUY9eBoMQcOuBNSDECIBDx4iuEZIQpidJIhsw9mesWw5Ee8+CtePKh48SD4N04eB40WNBRV3XR3eZEUGm37y0rNzS8sLqWXMyura+sb2c2tGx3GioPLQxmqW49pkCIAFwVKuI0UMN+TUP65yO/dgdKizC4xkETZ91A9ERnKGRWtliA+EeE1AqVMN85YA2egyTynCfnlHncGKWLyvuL6+VzdkFewz6lzhTkiNTVFvZj0Y75LEPAXLJtK47doTNhCkUXMIw04g1RIz3WRfqhgbMB91Mxt8N6Z5R2rQTKlMB0rH6cyJhvtYD3zOdPsOenvVG4n9ePcbOSTMRQRQjBHyqBNLiEdRUXbQgFHOTCEcSXMrZT3mGIcTaAZE4Iz+/Jf4h4VTgvOVTFXKk/TSJMdskvyxCHpEQuSJW4hJMH8kReyKv1aD1b9b7pDVlTWe2yS9Yn9F6BW</latexit> <latexit sha1_base64="KRxJjxRAFBSumCLgm+mSm7rf7k=">ACHicbVDLSgNBEJyNrxhfUY9eBoMQcOuBNSDECIBDx4iuEZIQpidJIhsw9mesWw5Ee8+CtePKh48SD4N04eB40WNBRV3XR3eZEUGm37y0rNzS8sLqWXMyura+sb2c2tGx3GioPLQxmqW49pkCIAFwVKuI0UMN+TUP65yO/dgdKizC4xkETZ91A9ERnKGRWtliA+EeE1AqVMN85YA2egyTynCfnlHncGKWLyvuL6+VzdkFewz6lzhTkiNTVFvZj0Y75LEPAXLJtK47doTNhCkUXMIw04g1RIz3WRfqhgbMB91Mxt8N6Z5R2rQTKlMB0rH6cyJhvtYD3zOdPsOenvVG4n9ePcbOSTMRQRQjBHyqBNLiEdRUXbQgFHOTCEcSXMrZT3mGIcTaAZE4Iz+/Jf4h4VTgvOVTFXKk/TSJMdskvyxCHpEQuSJW4hJMH8kReyKv1aD1b9b7pDVlTWe2yS9Yn9F6BW</latexit> Error • Generate an output 
 ˆ E P ( ˜ E = argmax ˜ E | F ) • Calculate its "badness" (e.g. 1-BLEU, 1-METEOR) 
 error( E, ˆ E ) = 1 − BLEU( E, ˆ E ) • We would like to minimize error

  7. Problem: Argmax is Non- differentiable • The argmax function makes discrete zero-one decisions • The gradient of this function is zero almost everywhere, not-conducive to gradient-based training

  8. <latexit sha1_base64="iwD7OmBG4KhDZEWl5K36ziE3oIk=">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</latexit> <latexit sha1_base64="iwD7OmBG4KhDZEWl5K36ziE3oIk=">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</latexit> <latexit sha1_base64="iwD7OmBG4KhDZEWl5K36ziE3oIk=">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</latexit> Risk • Risk is defined as the expected error P ( ˜ E | F ; θ )error( E, ˜ X risk( F, E, θ ) = E ) . ˜ E • This is includes the probability in the objective function! • Differentiable, but the sum is intractable • Minimum risk training minimizes risk, Shen et al. (2016) do so for NMT

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

Risk Management Workshop 1 Risk management workshop Why do we Risk Risk and need risk

Risk Management Workshop 1 Risk management workshop Why do we Risk Risk and need risk

FREE Lifelong Learning Event for Fasset Members Risk Management Workshop 1 Risk management workshop Why do we Risk Risk and need risk assessment control matrix management process Governance Risk appetite Agenda for and risk Risk

1.28k views • 105 slides
Risk Assessment Reduce the workers and Biosafety is an inexact science, and

Risk Assessment Reduce the workers and Biosafety is an inexact science, and

Risk Analysis Biosafety Risk assessment, Risk Risk analysis encom passes Management &amp; risk risk assessm ent, risk communication m anagem ent, and risk com m unication. Ephy Khaemba International Livestock Research Institute

506 views • 11 slides
Introduction to Machine Learning CMU-10701 10. Risk Minimization Barnabs Pczos 10. Risk

Introduction to Machine Learning CMU-10701 10. Risk Minimization Barnabs Pczos 10. Risk

Introduction to Machine Learning CMU-10701 10. Risk Minimization Barnabs Pczos 10. Risk Minimization 2 What have we seen so far? Several algorithms that seem to work fine on training datasets: Linear regression Nave Bayes

1.13k views • 51 slides
To risk or not to risk? Dr Iraklis Lazakis Dpt of NAOME University of Strathclyde LNG Bunkering

To risk or not to risk? Dr Iraklis Lazakis Dpt of NAOME University of Strathclyde LNG Bunkering

To risk or not to risk? Dr Iraklis Lazakis Dpt of NAOME University of Strathclyde LNG Bunkering &amp; Training Challenges LNG Bunkering &amp; Training Challenges To risk or not to risk? Risk, Reliability and criticality analysis

526 views • 30 slides
Minimum Bayes-Risk Methods in Automatic Speech Recognition Vaibhava Goel IBM William Byrne

Minimum Bayes-Risk Methods in Automatic Speech Recognition Vaibhava Goel IBM William Byrne

Minimum Bayes-Risk Methods in Automatic Speech Recognition Vaibhava Goel IBM William Byrne Johns Hopkins University Pattern Recognition in Speech and Language Processing Chap2 Outline Minimum Bayes-Risk Classification Framework

655 views • 61 slides
State At-Risk Training Presentation Alabama Department of Education Prevention and Support

State At-Risk Training Presentation Alabama Department of Education Prevention and Support

State At-Risk Training Presentation Alabama Department of Education Prevention and Support Services Section Telephone: (334) 242-8165 Fax: (334) 242-5962 State At-Risk Training Presentation Agenda State At-Risk (At-Risk) Overview

247 views • 13 slides
Training Scuba Divers: A Fatality Training Scuba Divers: A Fatality and Risk Analysis and Risk

Training Scuba Divers: A Fatality Training Scuba Divers: A Fatality and Risk Analysis and Risk

Training Scuba Divers: A Fatality Training Scuba Divers: A Fatality and Risk Analysis and Risk Analysis DAN Diving Fatality Workshop 8 10 April, 2010 Durham Hotel Durham, North Carolina, USA Dr. Drew Richardson President and Chief Operating

447 views • 44 slides
Market Risk Management Gen Ins. Risk Management Life Ins. Risk Management Op.

Market Risk Management Gen Ins. Risk Management Life Ins. Risk Management Op.

Market Risk Management Gen Ins. Risk Management Life Ins. Risk Management Op. Risk Management Board of Enterprise Risk Senior Management Committees Directors Management Credit Risk Cross-Border Exposure Country Rating

453 views • 6 slides
SESSION 4: WHAT IS RISK? Risk is ubiquitous and has always been around Risk has always been

SESSION 4: WHAT IS RISK? Risk is ubiquitous and has always been around Risk has always been

Aswath Damodaran 0 SESSION 4: WHAT IS RISK? Risk is ubiquitous and has always been around Risk has always been part of human existence. In our earliest days, the primary risks were physical and were correlated with material reward.

399 views • 11 slides
Learning Objectives E xpla in wha t QPR me a ns I de ntify risk fa c to rs a nd e a rly

Learning Objectives E xpla in wha t QPR me a ns I de ntify risk fa c to rs a nd e a rly

QPR Staff suicide prevention training Na me T itle / F a c ility e ma il Learning Objectives E xpla in wha t QPR me a ns I de ntify risk fa c to rs a nd e a rly wa rning sig ns o f suic ide Apply QPR princ iple

633 views • 40 slides
1 Risk in hospitals Legal Risk Clinical risk Risk arising out of failure to comply with

1 Risk in hospitals Legal Risk Clinical risk Risk arising out of failure to comply with

Effective Risk Management Dr Rohini Sridhar MD FRCPA Dr. Rohini Sridhar, MD, FRCPA Chief Operating Officer, Apollo Speciality Hospital, Madurai Risk Management programme Identification of risk Strategies to avoid risk Strategies

494 views • 8 slides
On the nature of financial risk: Why risk is so hard to measure and why risk models fail so often

On the nature of financial risk: Why risk is so hard to measure and why risk models fail so often

Case study Model risk Measure risk Nature of risk Minsky Conclusion On the nature of financial risk: Why risk is so hard to measure and why risk models fail so often J on Dan elsson Systemic Risk Centre London School of Economics

1.12k views • 89 slides
Preparedness Preliminary risk management Risk assessment Risk management Risk

Preparedness Preliminary risk management Risk assessment Risk management Risk

FAO/WHO guide for application of risk analysis principles to food safety emergencies Food Recall and Traceability (February 2013, Thailand) Preparedness Preliminary risk management Risk assessment Risk management Risk

440 views • 39 slides
Risk How to assess risk? Javier Estrada No universal agreement about it IESE Business

Risk How to assess risk? Javier Estrada No universal agreement about it IESE Business

Emerging Markets (II): Risk Javier Estrada ADFIN Winter/2014 1. Risk in EMs Introductory issues Wrong focus and implications Political risk: Assessment and incorporation 2. Where to Account for Risk? Possibilities

637 views • 10 slides
Mechanisms of Systemic Risk Contagion, Reinforcement, Redistribution Frank Schweitzer

Mechanisms of Systemic Risk Contagion, Reinforcement, Redistribution Frank Schweitzer

Mechanisms of Systemic Risk Frank Schweitzer ZIF Workshop Bielefeld, Germany 01-11 September 2009 1 / 36 Mechanisms of Systemic Risk Contagion, Reinforcement, Redistribution Frank Schweitzer fschweitzer@ethz.ch in collaboration with:

526 views • 48 slides
INTRODUCTION &amp; OVERVIEW LTC, Inc. Risk Adjustment Data Training August 2007 I/2-1

INTRODUCTION &amp; OVERVIEW LTC, Inc. Risk Adjustment Data Training August 2007 I/2-1

2007 REGIONAL TRAINING Risk Adjustment Data INTRODUCTION &amp; OVERVIEW LTC, Inc. Risk Adjustment Data Training August 2007 I/2-1 Purpose To provide participants new to risk adjustment the support needed to improve the quality and

1.37k views • 105 slides
L101: Incremental structured prediction Structured prediction reminder Given an input x (e.g. a

L101: Incremental structured prediction Structured prediction reminder Given an input x (e.g. a

L101: Incremental structured prediction Structured prediction reminder Given an input x (e.g. a sentence) predict y (e.g. a PoS tag sequence, cf lecture 6): Where Y is rather large and often depends on the input (e.g. L | x | in PoS tagging)

511 views • 22 slides
Incremental Classification: First Step into Lifelong Learning PAN Xinyu MMLab, Department of IE

Incremental Classification: First Step into Lifelong Learning PAN Xinyu MMLab, Department of IE

Incremental Classification: First Step into Lifelong Learning PAN Xinyu MMLab, Department of IE Multi-task Incremental Classification: Setup Training Data ... Target Model Multi-task Incremental Classification: Setup Training

746 views • 33 slides
How to Train Your Perceptron 16-385 Computer Vision (Kris Kitani) Carnegie Mellon University

How to Train Your Perceptron 16-385 Computer Vision (Kris Kitani) Carnegie Mellon University

PERCEPTRON How to Train Your Perceptron 16-385 Computer Vision (Kris Kitani) Carnegie Mellon University Lets start easy worlds smallest perceptron! w f y x y = wx (a.k.a. line equation, linear regression) Learning a Perceptron

656 views • 20 slides
A constructive approach to incremental learning Mario Rosario Guarracino October 12, 2006

A constructive approach to incremental learning Mario Rosario Guarracino October 12, 2006

High Performance Computing and Networking Institute National Research Council, Italy The Data Reference Model: A constructive approach to incremental learning Mario Rosario Guarracino October 12, 2006 10/13/2006 10:26 PM Acknowledgements

762 views • 41 slides
Learning Agent Learning Agents An Agent that observes its performance and adapts its

Learning Agent Learning Agents An Agent that observes its performance and adapts its

Learning Agent Learning Agents An Agent that observes its performance and adapts its decision-making to improve its performance in the future. MSE 2400 EaLiCaRA Dr. Tom Way MSE 2400 Evolution &amp; Learning 2 Agent Simple Reflex Agent

292 views • 4 slides
CSE 373: Topological Sort and Minimum Questions such that no vertex appears before another vertex

CSE 373: Topological Sort and Minimum Questions such that no vertex appears before another vertex

CSE 373: Topological Sort and Minimum Questions such that no vertex appears before another vertex that has an edge to it. Example applications: Spanning Trees In general: taking a dependency graph and coming up with order of execution. 3

54 views • 4 slides
CSE 326: Data Structures Graph representations Graphs Topological Sort Topological

CSE 326: Data Structures Graph representations Graphs Topological Sort Topological

Agenda Basic graph terminology CSE 326: Data Structures Graph representations Graphs Topological Sort Topological sort Reference: Weiss, Ch. 9 Hal Perkins Spring 2007 Lectures 22-23 2 Some Applications: Graph ADT?

332 views • 6 slides
Graphs-Topological Sort November 9, 2016 CMPE 250 Graphs-Topological Sort November 9, 2016 1 /

Graphs-Topological Sort November 9, 2016 CMPE 250 Graphs-Topological Sort November 9, 2016 1 /

Graphs-Topological Sort November 9, 2016 CMPE 250 Graphs-Topological Sort November 9, 2016 1 / 14 Introduction There are many problems involving a set of tasks in which some of the tasks must be done before others. For example, consider the

642 views • 14 slides

More recommend