OPEN HOUSE on Multi-Task and Complex Outputs Learning

Top-down vs. bottom-up methods for hierarchical classification

author: Claudio Gentile, Universitá dell´ Insubria

Description

We deal with hierarchical classification in the general case when an instance could be associated with multiple and/or partial paths in a given taxonomy. We approach the problem from different perspectives: ”top-down vs. bottom-up”, but also ”on-line vs. batch” and ”theoretical vs. experimental”. In this talk, we briefly present our recent research experience on this subject matter.

Categories

Top: Computer Science: Machine Learning: Structured Output

You might be experiencing some problems with Your Video player.

*Slides*
0:01	Top-down vs. bottom-up methods
0:38	State of the art:
3:06	Hierarchical classication model/1
6:40	Hierarchical classication model/2
11:45	Hierarchical classication model/3
16:34	Hierarchical classication model/4
18:59	Hierarchical classication model/4
20:00	Hierarchical classication model/4
20:58	Hierarchical classication model/4
21:08	B-SVM algorithm (bottom-up)/1
25:30	B-SVM algorithm (bottom-up)/2
29:44	B-SVM algorithm (bottom-up)/3:
30:07	H-SVM algorithm (top-down) [HCC03,CGZ06], ...
31:36	Experimental results/1: Datasets
34:25	Experimental results/2: Avg test error (H-loss)
36:05	Hierarchical classication model/3
36:35	Experimental results/2: Avg test error (H-loss)
37:08	Experimental results/3:
37:35	Hierarchical classication model/3
38:10	Experimental results/2: Avg test error (H-loss)
39:15	gExperimental results/3:
40:27	Experimental results/4: Breakdown on levels
40:51	Experimental results/5: Comments
41:22	On-line hierarchical classication [CGZ06]/1
42:41	On-line hierarchical classication/2
44:02	On-line hierarchical classication/3
44:12	On-line hierarchical classication [CGZ06]/1
44:19	On-line hierarchical classication/3
46:08	On-line hierarchical classication/4: Algorithm/1
46:58	On-line hierarchical classication/4: Algorithm/2
48:21	On-line hierarchical classication/5
50:50	Conclusions
51:48	Open questions
56:55	B-SVM algorithm (bottom-up)/1
63:25	B-SVM algorithm (bottom-up)/2

Lecture rating

People found this lecture:
Worth seeing
because it is:
Valuable and informative
Well presented
Easily understandable
Acceptably recorded
You need to login to cast your vote.

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.