The 18th European Conference on Machine Learning (ECML) and the 11th European Conference on Principles and Practice of Knowledge Discovery in Databases (PKDD)

Learning to Classify Documents with Only a Small Positive Training Set

author: Xiao-Li Li, Division of Information Systems, School of Computer Engineering, Nanyang Technological University

You might be experiencing some problems with Your Video player.

*Slides*
0:00	Learning to Classify Documents with Only a Small Positive Training Set
0:21	Outline
0:47	1. Introduction
1:14	Positive-Unlabeled (PU) Learning
2:11	PU Learning
2:38	An illustration of the typical PU Learning
3:10	Applications of the problem
3:44	Related works (1)
4:25	Related works (2)
4:48	Can we use the current techniques in some real applications?
4:56	A real-life business intelligence application - searching for information on related products
5:33	Current techniques can not work well!
5:40	The Assumption (1) of current techniques
6:00	Current Assumption (1)
6:17	PU learning with a small positive training set
7:13	The Assumption (2) of current techniques
7:47	2. The proposed techniques: Ideas
8:32	The proposed techniques: Ideas (Cont.)
8:48	The proposed techniques: LPLP
9:31	Step1: Selecting a set of representative word features from P
10:29	Select representative word features from P
10:47	Step2: identifying LP from U and probabilistically labeling the documents in LP
12:03	Identifying likely positives
12:40	The Naïve Bayesian method
13:09	Step3: EM algorithm
14:11	Build a final classifier
14:37	3. EMPIRICAL EVALUATION
15:50	Experiment setting
15:52	Performance of LPLP with different numbers of positive documents
16:18	LP + P or LP only?
16:48	The number of the representative features
17:12	Performance of LPLP, Roc-SVM and PEBL (using either P or LP) when using all positive
17:56	Comparative results when the number of the positive documents is small
18:08	Conclusions
18:29	Conclusions (cont.)

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.