Transcript Learning, Logic, and Probability: A Unified View

10-803
Markov Logic Networks
Instructor:
Pedro Domingos
Logistics

Instructor: Pedro Domingos






Email: [email protected]
Office: Wean 5317
Office hours: Thursdays 2:00-3:00
Course secretary: Sharon Cavlovich
Web: http://www.cs.washington.edu/homes/
pedrod/803/
Mailing list: [email protected]
Source Materials




Textbook:
P. Domingos & D. Lowd,
Markov Logic: An Interface Layer for AI,
Morgan & Claypool, 2008
Papers
Software:
Alchemy (alchemy.cs.washington.edu)
MLNs, datasets, etc.:
Alchemy Web site
Project

Possible projects:




Apply MLNs to problem you’re interested in
Develop new MLN algorithms
Other
Key dates/deliverables:





This week: Download Alchemy and start playing
October 9 (preferably earlier): Project proposal
November 6: Progress report
December 4: Final report and short presentation
Winter 2009: Conference submission (!)
What Is Markov Logic?


A unified language for AI/ML
Special cases:






First-order logic
Probabilistic models
Syntax: Weighted first-order formulas
Semantics: Templates for Markov nets
Inference: Logical and probabilistic
Learning: Statistical and ILP
Why Take this Class?

Powerful set of conceptual tools


Powerful set of software tools*



New way to look at AI/ML
Increase your productivity
Attempt more ambitious applications
Powerful platform for developing new
learning and inference algorithms

Many fascinating research problems
* Caveat: Not mature!
Sample Applications






Information extraction
Entity resolution
Link prediction
Collective classification
Web mining
Natural language
processing







Computational biology
Social network analysis
Robot mapping
Activity recognition
Personal assistants
Probabilistic KBs
Etc.
Overview of the Class






Background
Markov logic
Inference
Learning
Extensions
Your projects
Background

Markov networks




Representation
Inference
Learning
First-order logic



Representation
Inference
Learning (a.k.a. inductive logic programming)
Markov Logic




Representation
Properties
Relation to first-order logic and statistical
models
Related approaches
Inference





Basic MAP and conditional inference
The MC-SAT algorithm
Knowledge-based model construction
Lazy inference
Lifted inference
Learning

Weight learning






Generative
Discriminative
Incomplete data
Structure learning and theory revision
Statistical predicate invention
Transfer learning
Extensions




Continuous domains
Infinite domains
Recursive MLNs
Relational decision theory
Your Projects
(TBA)
Class begins here.
AI: The First 100 Years
IQ
Human
Intelligence
Artificial
Intelligence
1956
2006
2056
AI: The First 100 Years
IQ
Human
Intelligence
Artificial
Intelligence
1956
2006
2056
AI: The First 100 Years
IQ
1956
2006
Artificial
Intelligence
Human
Intelligence
2056
The Interface Layer
Applications
Interface Layer
Infrastructure
Networking
WWW
Email
Applications
Interface Layer
Internet
Protocols
Infrastructure
Routers
Databases
ERP
CRM
Applications
OLTP
Interface Layer
Infrastructure
Relational Model
Transaction
Management
Query
Optimization
Programming Systems
Applications
Interface Layer
Programming
High-Level Languages
Compilers
Infrastructure
Code
Optimizers
Hardware
Applications
Computer-Aided
Chip Design
Interface Layer
VLSI Design
Infrastructure
VLSI modules
Architecture
Applications
Operating
Systems
Compilers
Interface Layer
Microprocessors
ALUs
Infrastructure
Buses
Operating Systems
Applications
Software
Interface Layer
Virtual machines
Infrastructure
Hardware
Human-Computer Interaction
Applications
Productivity Suites
Interface Layer Graphical User Interfaces
Infrastructure
Widget Toolkits
Artificial Intelligence
Planning
Robotics
Applications
NLP
Vision
Multi-Agent
Systems
Interface Layer
Representation
Inference
Infrastructure
Learning
Artificial Intelligence
Planning
Robotics
Applications
NLP
Vision
Interface Layer
Multi-Agent
Systems
First-Order Logic?
Representation
Inference
Infrastructure
Learning
Artificial Intelligence
Planning
Robotics
Applications
NLP
Vision
Interface Layer
Multi-Agent
Systems
Graphical Models?
Representation
Inference
Infrastructure
Learning
Logical and Statistical AI
Field
Logical
approach
Statistical
approach
Knowledge
representation
First-order logic Graphical models
Automated
reasoning
Satisfiability
testing
Markov chain
Monte Carlo
Machine learning Inductive logic
programming
Neural networks
Planning
Markov decision
processes
Classical
planning
Natural language Definite clause
grammars
processing
Prob. contextfree grammars
We Need to Unify the Two



The real world is complex and uncertain
Logic handles complexity
Probability handles uncertainty
Artificial Intelligence
Planning
Robotics
Applications
NLP
Multi-Agent
Systems
Vision
Interface Layer
Markov Logic
Representation
Inference
Infrastructure
Learning