ILP/MLG/SRL collocated International conferences/workshops on learning from relational, graph-based and probabilistic knowledge

A Tutorial on Logic-Based Approaches to SRL

author: James Cussens, Department of Computer Science, University of York

Description

The relations in Statistical Relational Learning are often expressed using first-order logic, leading to formalisms which combine both logical and probabilistic representations. In this talk I intend to explain the most important consequences of adopting a logical approach to SRL. Defining distributions over 'possible worlds' is a common theme to many such approaches. Two prominent logic-based formalisms - Markov logic networks and PRISM programs - will be used as exemplars. Although the talk is tutorial in nature, I hope to make it interesting to those already familiar with this area!

You might be experiencing some problems with Your Video player.

*Slides*
0:00	A tutorial on logic-based approaches to SRL
0:38	Overview
1:50	Making the connections
2:16	Propositional logic
2:19	Propositional formulae as zero-one factors
4:55	Propositional probabilistic models
5:07	Generalising propositional logic
6:56	Further examples
7:51	Weighted clauses
10:11	Bayesian networks
10:51	Inference in propositional probabilistic models
12:30	First-order logic
12:43	Characteristics of first-order logic
14:23	Factor representation of universally quantified formulae
15:14	First-order models
16:32	Inference in first-order logic
18:05	First-order probabilistic models (parfactors)
18:51	Quantifying over random variables
21:17	What sort of probability distribution is defined?
25:42	Lifted inference in first-order probabilistic models - 1
27:09	Lifted inference in first-order probabilistic models - 2
27:24	Quantifying over random variables
27:47	Lifted inference in first-order probabilistic models - 2
28:40	Markov logic parfactors
29:52	Markov logic distribution
31:26	What’s the data?
32:41	First-order probabilistic models (generative)
33:04	Dynamic probabilistic models
34:57	The PRISM approach
35:44	An example "base" probability distribution
37:03	Defining a "base" distribution in PRISM
37:36	A joint instantiation determines a logical theory
38:33	Using a fixed, arbitrary logical theory to extend a base distribution
40:37	Working with target predicates
41:46	Computing target probabilities from a PRISM distribution
42:36	Abduction: A HMM example - 1
43:25	Abduction: A HMM example - 2
43:30	Abduction: A HMM example - 3
43:31	Abduction: A HMM example - 4
43:32	Abduction: A HMM example - 5
43:44	Computing probabilities by abduction - 1
43:46	Computing probabilities by abduction - 2
44:48	Computing probabilities by abduction - 3
44:49	Computing probabilities by abduction - 4
44:50	Computing probabilities by abduction - 5
44:51	Computing probabilities by abduction - 6
45:37	What’s the data?
46:29	Bayesian network learning for pedigrees
49:35	Some genetics
50:45	The problem
51:59	Defining a joint probability distribution
53:12	Pedigree and auxiliary variables
54:07	Ordered genotype variables
54:54	Unordered genotype variables
55:10	An example possible world
55:27	Penalty for heterozygosity
56:03	Encoding population frequencies
56:30	Priors on pedigrees
57:03	Incorporating evidence
57:27	An simple example
57:48	A result
58:33	Another result
58:38	An simple example
58:43	Another result
59:17	- Questions

Lecture rating

People found this lecture:
Worth seeing
because it is:
Valuable and informative
Well presented
Easily understandable
Acceptably recorded
You need to login to cast your vote.

Report a problem or upload files

If you have found a problem with this lecture or would like to send us extra material, articles, exercises, etc., please use our ticket system to describe your request and upload the data.
Enter your e-mail into the 'Cc' field, and we will keep you updated with your request's status.