thesis we will never really understand learning until we
play

Thesis: We will never really understand learning until we build - PowerPoint PPT Presentation

Sep 01, 2022 •17 likes •427 views

Never Ending Language Learning Tom M. Mitchell Machine Learning Department Carnegie Mellon University Thesis: We will never really understand learning until we build machines that learn many different things, from years of diverse

  1. Never Ending Language Learning Tom M. Mitchell Machine Learning Department Carnegie Mellon University

  2. Thesis: We will never really understand learning until we build machines that • learn many different things, • from years of diverse experience, • in a staged, curricular fashion, • and become better learners over time.

  3. NELL: Never-Ending Language Learner The task: • run 24x7, forever • each day: 1. extract more facts from the web to populate the ontology 2. learn to read (perform #1) better than yesterday Inputs: • initial ontology (categories and relations) • dozen examples of each ontology predicate • the web • occasional interaction with human trainers

  4. NELL today Running 24x7, since January, 12, 2010 Result: • KB with ~120 million confidence-weighted beliefs • learning to read • learning to reason • extending ontology

  5. NELL knowledge fragment football uses * including only correct beliefs equipment climbing skates helmet Canada Sunnybrook Miller uses equipment city hospital Wilson company country hockey Detroit GM politician CFRB radio Pearson Toronto hometown play hired competes airport home town with Stanley city Maple Leafs Red Cup company city won won Wings Toyota stadium team stadium league league Connaught city acquired paper city Air Canada NHL member created stadium Hino Centre plays in economic sector Globe and Mail Sundin Prius writer automobile Toskala Skydome Corrola Milson

  6. [Mitchell et al., CACM 2017] Improving Over Time Never Ending Language Learner 10’s of millions of beliefs reading skill tens of millions of beliefs à mean avg precision à 2010 time à 2016 2010 time à 2017

  7. Semi-Supervised Bootstrap Learning Learn which it ’ s underconstrained!! noun phrases are cities: San Francisco anxiety Paris Berlin selfishness Pittsburgh denial London Seattle Montpelier mayor of arg1 arg1 is home of live in arg1 traits such as arg1

  8. Key Idea 1: Coupled semi-supervised training: multi-view and multi-task Y: person f: X à Y X: noun phrase hard (underconstrained) semi-supervised learning

  9. Key Idea 1: Coupled semi-supervised training: multi-view and multi-task person sport athlete coach team Y: person f: X à Y noun phrase noun phrase noun phrase text context morphology URL specific X: noun phrase “ __ is my son ” ends in ‘ … ski’ appears in list2 at URL35401 hard much easier (underconstrained) (more constrained) semi-supervised semi-supervised learning learning

  10. Supervised training of 1 function : y : person x :

  11. Coupled training of 2 functions : y : person x :

  12. NELL Learned Contexts for “Hotel” (~1% of total) "_ is the only five-star hotel” "_ is the only hotel” "_ is the perfect accommodation" "_ is the perfect address” "_ is the perfect lodging” "_ is the sister hotel” "_ is the ultimate hotel" "_ is the value choice” "_ is uniquely situated in” "_ is Walking Distance” "_ is wonderfully situated in” "_ las vegas hotel” "_ los angeles hotels” "_ Make an online hotel reservation” "_ makes a great home-base” "_ mentions Downtown” "_ mette a disposizione” "_ miami south beach” "_ minded traveler” "_ mucha prague Map Hotel” "_ n'est qu'quelques minutes” "_ naturally has a pool” "_ is the perfect central location” "_ is the perfect extended stay hotel” "_ is the perfect headquarters” "_ is the perfect home base” "_ is the perfect lodging choice" "_ north reddington beach” "_ now offer guests” "_ now offers guests” "_ occupies a privileged location” "_ occupies an ideal location” "_ offer a king bed” "_ offer a large bedroom” "_ offer a master bedroom” "_ offer a refrigerator” "_ offer a separate living area" "_ offer a separate living room” "_ offer comfortable rooms” "_ offer complimentary shuttle service” "_ offer deluxe accommodations” "_ offer family rooms” "_ offer secure online reservations” "_ offer upscale amenities” "_ offering a complimentary continental breakfast” "_ offering comfortable rooms” "_ offering convenient access” "_ offering great lodging” "_ offering luxury accommodation” "_ offering world class facilities” "_ offers a business center" "_ offers a business centre” "_ offers a casual elegance” "_ offers a central location” “_ surrounds travelers” …

  13. NELL Highest Weighted* string fragments: “Hotel” 1.82307 SUFFIX=tel 1.81727 SUFFIX=otel 1.43756 LAST_WORD=inn 1.12796 PREFIX=in 1.12714 PREFIX=hote 1.08925 PREFIX=hot 1.06683 SUFFIX=odge 1.04524 SUFFIX=uites 1.04476 FIRST_WORD=hilton 1.04229 PREFIX=resor 1.02291 SUFFIX=ort 1.00765 FIRST_WORD=the 0.97019 SUFFIX=ites 0.95585 FIRST_WORD=le 0.95574 PREFIX=marr 0.95354 PREFIX=marri 0.93224 PREFIX=hyat 0.92353 SUFFIX=yatt 0.88297 SUFFIX=riott 0.88023 PREFIX=west * logistic regression 0.87944 SUFFIX=iott

  14. Type 1 Coupling: Co-Training, Multi-View Learning Theorem (Blum & Mitchell, 1998) : y : person If f 1 ,and f 2 are PAC learnable from noisy labeled data, and X 1 , X 2 are conditionally independent given Y, Then f 1 , f 2 are PAC learnable from polynomial unlabeled data plus a weak initial predictor x :

  15. Type 1 Coupling: Co-Training, Multi-View Learning [Blum & Mitchell; 98] [Dasgupta et al; 01 ] [Balcan & Blum; 08] [Ganchev et al., 08] y : person [Sridharan & Kakade, 08] [Wang & Zhou, ICML10] x :

  16. Type 1 Coupling: Co-Training, Multi-View Learning [Blum & Mitchell; 98] sample complexity drops exponentially [Dasgupta et al; 01 ] in the number of views of X [Balcan & Blum; 08] [Ganchev et al., 08] y : person [Sridharan & Kakade, 08] [Wang & Zhou, ICML10] x :

  17. Type 2 Coupling: Multi-task, Structured Outputs [Daume, 2008] [Bakhir et al., eds. 2007] [Roth et al., 2008] [Taskar et al., 2009] person sport [Carlson et al., 2009] athlete coach team subset/superset athlete(NP) à person(NP) NP mutual exclusion athlete(NP) à NOT sport(NP) sport(NP) à NOT athlete(NP)

  18. Multi-view, Multi-Task Coupling person sport athlete coach team NP text NP NP HTML NP : context morphology contexts distribution

  19. Type 3 Coupling: Relations and Argument Types playsSport(a,s) coachesTeam(c,t) playsForTeam(a,t) teamPlaysSport(t,s) NP1 NP2

  20. Type 3 Coupling: Relations and Argument Types playsSport(a,s) coachesTeam(c,t) playsForTeam(a,t) teamPlaysSport(t,s) person sport person sport athlete athlete team coach team coach NP1 NP2

  21. Type 3 Coupling: Relations and Argument Types playsSport(NP1,NP2) à athlete(NP1), sport(NP2) playsSport(a,s) coachesTeam(c,t) playsForTeam(a,t) teamPlaysSport(t,s) person sport person sport athlete athlete team coach team coach NP1 NP2

  22. Type 3 Coupling: Relations and Argument Types over 4000 coupled functions in NELL playsSport(a,s) coachesTeam(c,t) playsForTeam(a,t) teamPlaysSport(t,s) person sport person sport athlete athlete team coach team coach NP1 1 NP1 2 NP2 1 NP2 multi-view consistency subset/superset argument type consistency mutual exclusion

  23. How to train approximation to EM: • E step: predict beliefs from unlabeled data (ie., the KB) • M step: retrain NELL approximation: • bound number of new beliefs per iteration, per predicate • rely on multiple iterations for information to propagate, partly through joint assignment, partly through training examples Better approximation: • Joint assignments based on probabilistic soft logic [Pujara, et al., 2013] [Platanios et al., 2017]

  24. If coupled learning is the key, how can we get new coupling constraints?

  25. Key Idea 2: Learn inference rules PRA: [Lao, Mitchell, Cohen, EMNLP 2011] competes economic If: x1 x2 x3 with sector (x1,x2) (x2, x3) Then: economic sector (x1, x3) with probability 0.9

  26. Key Idea 2: Learn inference rules PRA: [Lao, Mitchell, Cohen, EMNLP 2011] economic sector competes economic If: x1 x2 x3 with sector (x1,x2) (x2, x3) Then: economic sector (x1, x3) with probability 0.9

  27. Learned Rules are New Coupling Constraints! 0.93 playsSport(?x,?y) ß playsForTeam(?x,?z), teamPlaysSport(?z,?y) playsSport(a,s) coachesTeam(c,t) playsForTeam(a,t) teamPlaysSport(t,s) person sport person sport athlete athlete team coach team coach NP1 NP2

  28. Learned Rules are New Coupling Constraints! • Learning X makes one a better learner of Y • Learning Y makes one a better learner of X X = reading functions: text à beliefs Y = Horn clause rules: beliefs à beliefs

  29. Consistency and Correctness what is the relationship? under what conditions?

  30. The core problem: • Unsupervised agents can measure their internal consistency , but not their correctness Challenge: • Under what conditions does consistency à correctness ?

  31. [Platanios, Blum, Mitchell] Problem setting: • have N different estimates of target function = NELL category “city” = classifier based on i th view of = noun phrase

Recommend

Group-6 Assignment IN5480 Shuvo Mahmuda Main Motivations Understand HCI and Machine Learning

Group-6 Assignment IN5480 Shuvo Mahmuda Main Motivations Understand HCI and Machine Learning

Group-6 Assignment IN5480 Shuvo Mahmuda Main Motivations Understand HCI and Machine Learning from an different Angles Master thesis in Machine Learning and AI Questions suggested by the Group How to tackle ambiguity in AI and Human

425 views • 10 slides
HONORS THESIS PRESENTATION GUIDELINES FOR THESIS ADVISORS AND SECOND READERS Thesis Presentation :

HONORS THESIS PRESENTATION GUIDELINES FOR THESIS ADVISORS AND SECOND READERS Thesis Presentation :

HONORS THESIS PRESENTATION GUIDELINES FOR THESIS ADVISORS AND SECOND READERS Thesis Presentation : The culmination of the senior thesis is a 30-minute thesis presentation with the student, faculty thesis advisor and second reader in attendance.

473 views • 4 slides
Introduction to Machine Learning k -Nearest Neighbors Regression Learning goals Understand the

Introduction to Machine Learning k -Nearest Neighbors Regression Learning goals Understand the

Introduction to Machine Learning k -Nearest Neighbors Regression Learning goals Understand the basic idea of k-NN Know different distance measures for different scales of feature variables Understand that k-NN has no optimization step

569 views • 13 slides
The Frontier Thesis: How & Why the Riverina Was Won The Frontier Thesis The Frontier Thesis:

The Frontier Thesis: How & Why the Riverina Was Won The Frontier Thesis The Frontier Thesis:

The Frontier Thesis: How & Why the Riverina Was Won The Frontier Thesis The Frontier Thesis: How & Why the Riverina Was Won Dr Rodney Gillett AFL 150 th History Conference July 2008 Victoria University Melbourne The Frontier Thesis

530 views • 26 slides
Learning Outcomes I understand the sources of capacitance in CMOS circuits I understand how

Learning Outcomes I understand the sources of capacitance in CMOS circuits I understand how

2-7.1 2-7.2 Learning Outcomes I understand the sources of capacitance in CMOS circuits I understand how delay scales with resistance, capacitance and voltage Spiral 2-7 I can determine appropriate width of PMOS and NMOS transistors

384 views • 10 slides
Learning from Limited Labeled Data (but a lot of unlabeled data) NELL as a case study Tom M.

Learning from Limited Labeled Data (but a lot of unlabeled data) NELL as a case study Tom M.

Learning from Limited Labeled Data (but a lot of unlabeled data) NELL as a case study Tom M. Mitchell Carnegie Mellon University Thesis: We will never really understand learning until we build machines that learn many different things,

582 views • 43 slides
Introduction to Machine Learning ML-Basics: Data Learning goals 10 Understand structure of

Introduction to Machine Learning ML-Basics: Data Learning goals 10 Understand structure of

Introduction to Machine Learning ML-Basics: Data Learning goals 10 Understand structure of tabular data in ML 5 Understand difference between 0 y target and features 5 Understand difference between labeled and unlabeled data 10

670 views • 9 slides
Thesis | Biophilic Design CDES 4721 Sarah Lafrinere Thesis & Exhibition Lisa Abendroth

Thesis | Biophilic Design CDES 4721 Sarah Lafrinere Thesis & Exhibition Lisa Abendroth

Thesis | Biophilic Design CDES 4721 Sarah Lafrinere Thesis & Exhibition Lisa Abendroth Thesis Question | How can Biophilic Design be used in higher education to increase concentration and build a productive learning environment for

253 views • 7 slides
Learning objectives Understand the purposes of planning and monitoring Distinguish

Learning objectives Understand the purposes of planning and monitoring Distinguish

Learning objectives Understand the purposes of planning and monitoring Distinguish strategies from plans, and Planning and Monitoring the Process understand their relation Understand the role of risks in planning Understand

190 views • 16 slides
Dementia & Learning Disability Standard 1 Learning outcomes 9.1 Understand the experiences

Dementia & Learning Disability Standard 1 Learning outcomes 9.1 Understand the experiences

The CARE CERTIFICATE Awareness of Mental Health, Dementia & Learning Disability Standard 1 Learning outcomes 9.1 Understand the experiences of people with mental health needs, dementia or learning disabilities 9.2 Understand the

304 views • 26 slides
Spiral 2-5 Sequential Logic Constructs 2-5.2 Learning Outcomes I understand how a bistable

Spiral 2-5 Sequential Logic Constructs 2-5.2 Learning Outcomes I understand how a bistable

2-5.1 Spiral 2-5 Sequential Logic Constructs 2-5.2 Learning Outcomes I understand how a bistable works I understand how a bistable holds, sets, and resets I understand the issues that glitches pose to bistables and the need for

1.07k views • 69 slides
Honors Thesis & Thesis Presentation Guidelines for Thesis Advisers and Second Readers

Honors Thesis & Thesis Presentation Guidelines for Thesis Advisers and Second Readers

Honors Thesis & Thesis Presentation Guidelines for Thesis Advisers and Second Readers OVERVIEW OF THE HONORS THESIS All Honors students must complete a year-long Honors Thesis project which is due in the students last semester Upon

385 views • 4 slides
Spiral 3-1 Hardware/Software Interfacing 3-1.2 Learning Outcomes I understand the

Spiral 3-1 Hardware/Software Interfacing 3-1.2 Learning Outcomes I understand the

3-1.1 Spiral 3-1 Hardware/Software Interfacing 3-1.2 Learning Outcomes I understand the PicoBlaze bus interface signals: PORT_ID, IN_PORT, OUT_PORT, WRITE_STROBE I understand how a memory map provides the agreement between

829 views • 40 slides
How to write a Master's Thesis PhDr. Ing. Antonn Pavlek, Ph.D. PREPARATION Students

How to write a Master's Thesis PhDr. Ing. Antonn Pavlek, Ph.D. PREPARATION Students

How to write a Master's Thesis PhDr. Ing. Antonn Pavlek, Ph.D. PREPARATION Students learning how to write a Master's Thesis should first focus on a central thesis question it must be presented and subsequently answered. A

665 views • 17 slides
THESIS DEFENCE AND PRESENTATION RECOMMENDATIONS University of Veterinary Medicine and Pharmacy in

THESIS DEFENCE AND PRESENTATION RECOMMENDATIONS University of Veterinary Medicine and Pharmacy in

THESIS DEFENCE AND PRESENTATION RECOMMENDATIONS University of Veterinary Medicine and Pharmacy in Koice To enable the committee to identify content of the thesis quickly and accurately to understand quickly and accurately what you did and

1.46k views • 26 slides
Spiral 2-8 Cell Layout 2-8.2 Learning Outcomes I understand how a digital circuit is

Spiral 2-8 Cell Layout 2-8.2 Learning Outcomes I understand how a digital circuit is

2-8.1 Spiral 2-8 Cell Layout 2-8.2 Learning Outcomes I understand how a digital circuit is composed of layers of materials forming transistors and wires I understand how each layer is expressed as geometric patterns called masks I

532 views • 39 slides
Introduction to Machine Learning Polynomial Regression Models Learning goals Understand how to

Introduction to Machine Learning Polynomial Regression Models Learning goals Understand how to

Introduction to Machine Learning Polynomial Regression Models Learning goals Understand how to add flexibility to the linear model by using 2.0 f(x) for d = 1 (linear) f(x) for d = 5 1.5 polynomials f(x) for d = 25 1.0 0.5 y Understand

496 views • 8 slides
Learning Objectives At the end of this session, attendees will: Understand what issues

Learning Objectives At the end of this session, attendees will: Understand what issues

New Jersey Vaccine Preventable Disease Program Vaccine for Children Program Christopher Ryan, MICP, CPM Learning Objectives At the end of this session, attendees will: Understand what issues constitute fraud and abuse Understand best

356 views • 23 slides
Spiral 2-6 Semiconductor Material MOS Theory 2-6.2 Learning Outcomes I understand why a

Spiral 2-6 Semiconductor Material MOS Theory 2-6.2 Learning Outcomes I understand why a

2-6.1 Spiral 2-6 Semiconductor Material MOS Theory 2-6.2 Learning Outcomes I understand why a diode conducts current under forward bias but does not under reverse bias I understand the three modes of operation of a MOS transistor and

703 views • 55 slides
Learning Methods: Part 1 CS 760@UW-Madison Goals for the lecture you should understand the

Learning Methods: Part 1 CS 760@UW-Madison Goals for the lecture you should understand the

Evaluating Machine Learning Methods: Part 1 CS 760@UW-Madison Goals for the lecture you should understand the following concepts bias of an estimator learning curves stratified sampling cross validation confusion

684 views • 27 slides
LEARNING INTENTIONS To understand what is meant by the term NPS - New Psychoactive Substances

LEARNING INTENTIONS To understand what is meant by the term NPS - New Psychoactive Substances

New P sychoactiv e Substances (NPS) LEARNING INTENTIONS To understand what is meant by the term NPS - New Psychoactive Substances To understand the reasons why it is misleading to call them legal highs To become aware of what the

379 views • 20 slides
Learning objectives Understand the rationale for systematic (non- random) selection of test

Learning objectives Understand the rationale for systematic (non- random) selection of test

Learning objectives Understand the rationale for systematic (non- random) selection of test cases Understand the basic concept of partition testing Functional testing and its underlying assumptions Understand why functional test

220 views • 5 slides
Learning Methods: Part 2 CS 760@UW-Madison Goals for the last lecture you should understand the

Learning Methods: Part 2 CS 760@UW-Madison Goals for the last lecture you should understand the

Evaluating Machine Learning Methods: Part 2 CS 760@UW-Madison Goals for the last lecture you should understand the following concepts bias of an estimator learning curves stratified sampling cross validation

542 views • 27 slides
Learning Outcomes Understand what Academic Integrity is and why it is important to demonstrate

Learning Outcomes Understand what Academic Integrity is and why it is important to demonstrate

Learning Outcomes Understand what Academic Integrity is and why it is important to demonstrate is as: A Student enrolled at a college A Professional in any workplace. Recognize and avoid Academic Offences Understand the Academic

617 views • 37 slides

More recommend