Difference between revisions of "10-601 SSL"

Revision as of 16:08, 30 October 2013

You should know:

What collaborative filtering is.
How nearest-neighbor methods for CF work.
How to formulate CF as a regression or classification problem.
How matrix factorization can be used for CF.
How PCA, SVD, k-means, and other clustering methods relate to matrix factorization.

@@ Line 1: / Line 1: @@
 === Slides ===
-* [http://www.cs.cmu.edu/~wcohen/10-601/ssl.pptx Slides in PowerPoint].
+* [http://www.cs.cmu.edu/~wcohen/10-601/cf.pptx Slides in PowerPoint].
 === Readings ===
-* This is not covered in Mitchell.  An optional reading is an excellent short textbook by Jerry Zhu: [http://www.morganclaypool.com/doi/abs/10.2200/S00196ED1V01Y200906AIM006 Introduction to Semi-Supervised Learning Synthesis Lectures on Artificial Intelligence and Machine Learning], Chapters 1-3 + 5.  This is a free PDF if you're on the CMU network.
+* This is not covered in Mitchell.
 ===  Summary  ===
 You should know:
-* What semi-supervised learning is - i.e., what the inputs and outputs are.
+* What collaborative filtering is.
-* How K-means and mixture-models can be extended to perform SSL.
+* How nearest-neighbor methods for CF work.
-* The difference between transductive and inductive semi-supervised learning.
+* How to formulate CF as a regression or classification problem.
-* What graph-based SSL is.
+* How matrix factorization can be used for CF.
-* The definition/implementation of the harmonic function SSL method (variously called wvRN, HF, co-EM, ...)
+* How PCA, SVD, k-means, and other clustering methods relate to matrix factorization.