9.520-A | Spring 2001 | Graduate

Networks for Learning: Regression and Classification


  1. Two problem sets.
  2. There will be a term paper for graduate credit due on the last day of class. The paper should either present some new work addressing a relevant research problem, or provide a critical analysis of 2-3 papers on some aspect of learning, approximation, or networks. A short description of the project you intend to work on is due before Lec #18. Below is a list of available topics; other topics must be approved by the instructor.

List of available projects:

Project 1: Hypothesis Testing with Small Sets.

Project 2: Connection between MED and Regularization.

Project 3: Kernels for Strings.

Project 4: Feature Selection for SVMs: Theory and Experiments.

Project 5: Morphable Models and Roweis’ Nonlinear Dimensionality Reduction.

Project 6: Optimal Bayes Classification Rule and SVMs: Estimation of ROC Curves.

Project 7: IOHMMs: Evaluation of HMMs vs Direct Classifiers like SVMs.

Project 8: Reusing the Test Set: Datamining Bounds.

Project 9: Stability and Generalization.

Project 10: Learning with Very Large Dataset.

Project 11: Bagging, Boosting, and Stability.

Project 12: Local vs. Global Classifiers: Simulations and Use of Prior Knowledge.

Project 13: Invariance to Measure of the RKHS Norm in the Continuum.

Project 14: Concentration Experiments: Dot Products vs Distances in Very High Dimension.

Course Info

As Taught In
Spring 2001
Learning Resource Types
Problem Sets
Written Assignments