Jul 9, 2006 
18 videos 
Open House on Multi-Task and Complex Outputs Learning, London 2006
The Open House is part of the PASCAL's fifth thematic programme. This scientific programme brings together researchers around two themes:
- learning tasks where the target is complex (i.e. large number of different predictions) but has structure (sequence, tree, graph) that can be utilized for learning effectively;
- multi-task learning, where several dependent learning tasks are to be tackled at once.
In particular, we shall focus on the following research topics/issues.
Introductions
07:54
Introduction
Feb 25, 2007 5333 views
Lectures
56:25
Multitask learning: the Bayesian way
Feb 25, 2007 6032 views
24:56
How to Teach Support Vector Machine to Learn Vector Outputs
Feb 25, 2007 6543 views
20:40
XML structure mapping
Feb 25, 2007 5847 views
48:05
Estimation of gradients and coordinate covariation in classification
Feb 25, 2007 4425 views
01:04:00
Inductive transfer via embeddings into a common feature space
Feb 25, 2007 3372 views
01:03:55
Going beyond bag-of-words: dealing with a text as a graph of triples
Feb 25, 2007 8745 views
55:51
Output kernel tree
Feb 25, 2007 4582 views
50:24
Multi-task feature learning
Feb 25, 2007 7183 views
13:48
Learning Structured Outputs via Kernel Dependency Estimation and Stochastic Gram...
Feb 25, 2007 3458 views
29:59
Targeted PDF Learning
Feb 25, 2007 4880 views
31:47
Machine Learning for Sequential Data: A Comparative Study with Applications to N...
Feb 25, 2007 6940 views
36:15
Slow subspace learning from stationary processes
Feb 25, 2007 3139 views
01:20:01
Learning shared representations for object recognition
Feb 25, 2007 9686 views
47:57
Efficient max-margin Markov learning via conditional gradient and probabilistic ...
Feb 25, 2007 4343 views
50:05
Learning Nonparametric Priors from Multiple Tasks
Feb 25, 2007 4410 views
01:08:00
Top-down vs. bottom-up methods for hierarchical classification
Feb 25, 2007 8372 views
19:38
Learning with structured inputs
Feb 25, 2007 3404 views