Machine Learning Summer School 2006 - Taipei

Machine Learning Reductions

author: John Langford, Yahoo! Research

Description

There are several different classification problems commonly encountered in real world applications such as 'importance weighted classification', 'cost sensitive classification', 'reinforcement learning', 'regression' and others. Many of these problems can be related to each other by simple machines (reductions) that transform problems of one type into problems of another type. Finding a reduction from your problem to a more common problem allows the reuse of simple learning algorithms to solve relatively complex problems. It also induces an organization on learning problems — problems that can be easily reduced to each other are 'nearby' and problems which can not be so reduced are not close.

Categories

Top: Computer Science: Machine Learning

You might be experiencing some problems with Your Video player.

*Slides*
0:00	Machine Learning Reductions Tutorial
1:07	Scenario 1
2:24	Scenario 2
3:56	Where did the Hollywood ending go?
5:55	Characteristics of Learning Reductions
7:07	It's reductionist (= good research direction)
8:04	Elemental
11:37	It's easy (= you can use it too)
12:32	Given a Binary classifier, how can we solve
13:26	Classification Definition
16:20	Importance Weighted Classification
17:55	The core theorem: folklore
20:51	How do we change distributions?
23:03	Distribution Transform: Rejection Sampling
24:04	Costing (Sw, A)
24:59	Costing+classifier applied to the KDD-98 dataset
27:38	Costing+classifier applied to the DMEF2 dataset
28:12	Given a Binary classifier, how can we solve
29:27	Square Error Regression
31:11	Reasons for the Regression Problem
32:33	The Probing Method: Observations
35:37	The Probing Algorithm
37:47	The Probing Method: Details
39:17	Comparison for Probing with Squared Error
44:46	The one classifier trick
47:06	Probing Theory
51:06	The proof, pictorially
57:15	The proof, mathematically
60:17	Proof, continued
62:33	Proof II: Properties of most efficient error inducing method
64:14	An Modification: Quantile Regression
66:12	normalized performance
71:07	Some Caveats
73:20	Given a Binary classifier, how can we solve

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.