25th International Conference on Machine Learning (ICML), Helsinki 2008

Discriminative Structure and Parameter Learning for Markov Logic Networks

author: Tuyen Ngoc Huynh, The University of Texas at Austin
published: July 28, 2008, recorded: July 2008, views: 312

Slides

Slides
0:00	Discriminative Structure and Parameter Learning for Markov Logic Networks
0:22	Motivation
1:27	Background
1:30	Markov Logic Networks
2:48	Inference in MLNs
4:20	Existing Learning Methods for MLNs
5:37	Initial Results
6:52	Generative vs Discriminative in SRL
8:03	Proposed Approach - 1
8:06	Proposed Approach - 2
8:33	Discriminative Structure Learning - 1
10:03	Discriminative Structure Learning - 2
10:47	Discriminative Weight Learning
11:37	Exact Inference
12:15	L1-Regularization
14:05	L1 Weight Learner
14:34	Experiments
14:37	Data Sets
15:28	Methodology - 1
15:51	Methodology - 2
16:26	Methodology - 3
16:49	Methodology - 4
17:08	Methodology - 5
17:21	Methodology - 6
17:53	Methodology - 7
18:25	Methodology - 8
18:41	Conclusion
19:25	Thank you! Questions?
21:19	- Questions

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.

Lecture popularity: You need to login to cast your vote.

Description

Markov logic networks (MLNs) are an expressive representation for statistical relational learning that generalizes both first-order logic and graphical models. Existing methods for learning the logical structure of an MLN are not discriminative; however, many relational learning problems involve specific target predicates that must be inferred from given background information. We found that existing MLN methods perform very poorly on several such ILP benchmark problems, and we present improved discriminative methods for learning MLN clauses and weights that outperform existing MLN and traditional ILP methods.