en-de
en-es
en-fr
en-sl
en
en-zh
0.25
0.5
0.75
1.25
1.5
1.75
2
Efficient max-margin Markov learning via conditional gradient and probabilistic inference
Published on Feb 25, 20074327 Views
We present a general and efficient optimisation methodology for for max-margin sructured classification tasks. The efficiency of the method relies on the interplay of several techiques: marginalizatio
Related categories
Chapter list
Efficient Max-Margin Markov00:02
rStructured Multilabel Classi cation00:33
Hypergraph structure?02:07
Hierarchical Multilabel Classi cation:02:31
Classi cation on a DAG: Gene ontology prediction03:32
The classi cation model04:51
Max-margin Structured output learning06:11
Optimization problem07:34
Decomposable representations08:58
Orthogonal features10:25
Orthogonal features12:34
Additive features13:44
Loss functions15:31
Hierarchical losses18:12
Optimization problem20:23
Marginalized problem20:39
Marginalized problem22:03
Size of the marginal dual problem23:51
Decomposing the model25:18
Conditional Gradient method26:49
Conditional Gradient Ascent27:58
Conditional Gradient Ascent28:11
Conditional Gradient Ascent28:30
Conditional Gradient Ascent28:40
Conditional Gradient Ascent29:15
Decomposing the model29:20
Finding update directions eciently30:48
Finding update directions eciently31:32
Finding update directions eciently33:04
Solving the inference problem34:17
Conditional Gradient Ascent35:25
Solving the inference problem35:45
Experiments36:21
Optimization eciency37:44
Prediction accuracy: Levelwise F138:40
Scalability?41:17
Conclusions42:39
Future work and open problems44:11