0.25
0.5
0.75
1.25
1.5
1.75
2
Recent Advances in Large Linear Classification
Published on Mar 27, 20143537 Views
Linear classification is a useful tool in machine learning and data mining. For some data in a rich dimensional space, the prediction performance of linear classifiers has shown to be close to that of
Related categories
Chapter list
Recent Advances in Large-scale Linear Classification00:00
This talk00:40
Outline02:14
Outline: Introduction02:36
Linear and Nonlinear Classification02:37
Linear and Nonlinear Classification (Cont'd)03:35
Why Linear Classification05:50
Why Linear Classification (Cont'd)08:13
Linear is Useful in Some Places09:11
Comparison Between Linear and Nonlinear (Training Time & Testing Accuracy) - 110:45
Comparison Between Linear and Nonlinear (Training Time & Testing Accuracy) - 210:52
Comparison Between Linear and Nonlinear (Training Time & Testing Accuracy) - 312:05
Binary Linear Classification15:05
Loss Functions15:38
Loss Functions (Cont'd) - 218:01
Loss Functions (Cont'd) - 118:25
Outline: Optimization Methods18:44
Optimization: 2nd Order Methods25:01
Optimization: 2nd Order Methods (Cont'd)25:08
2nd-order Methods (Cont'd)25:18
Optimization: 1st Order Methods29:16
1st Order Methods (Cont'd) - 130:02
1st Order Methods (Cont'd) - 230:32
1st Order Methods (Cont'd) - 333:01
1st Order Methods (Cont'd) - 433:03
Comparisons33:04
Objective values (Time in Seconds)33:51
Analysis34:52
An Example When # Features Small36:07
Outline: Extension of Linear Classification37:23
Extension of Linear Classification37:28
Linear Methods to Explicitly Train Ø(xi)38:35
Example: Dependency Parsing40:26
Example: Dependency Parsing (Cont'd)41:36
Example: Classifier in a Small Device - 246:50
Example: Classifier in a Small Device - 348:09
Discussion49:11
Example: Classifier in a Small Device - 150:15
Outline: Discussion and Conclusions51:05
Big-data Linear Classification - 151:09
Big-data Linear Classification - 253:19
Conclusions54:41