6th International Workshop on Mining and Learning with Graphs (MLG), Helsinki 2008

Markov Logic Improves Protein β-Partners Prediction

author: Marco Lippi, Dipartimento di Sistemi e Informatica, University of Firenze
published: Aug. 25, 2008, recorded: July 2008, views: 139

Turn off the lights

You might be experiencing some problems with Your Video player.

Slides

Slides
0:00	Markov Logic Improves Protein β-Partners Prediction
0:16	3D Protein Structure Prediction
1:38	β-Partners: A Link Prediction Task - 1
2:25	β-Partners: A Link Prediction Task - 2
3:11	State-of-the-art Architecture: BetaPro
4:19	β-Partners: A Link Prediction Task - 3
5:53	Markov Logic - 1
6:35	Markov Logic - 2
7:26	Markov Logic - 3
8:06	Curse of Dimensionality
9:39	Expressivity of the Model
10:03	Markov Logic Networks with Grounding-Specific Weights - 1
10:53	Markov Logic Networks with Grounding-Specific Weights - 2
11:58	Markov Logic Networks with Grounding-Specific Weights - 3
13:03	Markov Logic Networks with Grounding-Specific Weights - 2
13:27	Markov Logic Networks with Grounding-Specific Weights - 3
13:28	The Data Set
14:34	The Model: Hard Rules
15:33	The Model: More Complex Rules - 1
17:08	The Model: More Complex Rules - 2
17:47	Plugging in BetaPro Probabilities
18:58	The Model: A Second-Stage MLN
20:07	Results Obtained on 10-Fold Cross Validation - 1
21:02	Results Obtained on 10-Fold Cross Validation - 2
21:54	Examples: Comparison with BetaPro - 1
23:18	Examples: Comparison with BetaPro - 3
23:45	Conclusions and Future Work
25:25	- 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

Protein β-partners prediction is an important problem in protein structure that can be naturally formulated as supervised link prediction. We show that prediction performance can be improved using a hybrid solution based on Markov logic networks with grounding-specific weights.