Lecture Notes

The lecture notes for this course were prepared by Alexander Rakhlin, a student in the class.


Lec # Topics
1 Introduction
2 Voting Classifiers, Training Error of Boosting (PDF)
3 Support Vector Machines (SVM) (PDF)
4 Generalization Error of SVM (PDF)
5 One Dimensional Concentration Inequalities; Bennett's Inequality (PDF)
6 Bernstein's Inequality (PDF)
7 Hoeffding's Inequality, Hoeffding-Chernoff Inequality (PDF)
8 Vapnik-Chervonenkis Classes of Sets (PDF)
9 Properties of VC Classes of Sets (PDF)
10 Symmetrization; Pessimistic VC Inequality (PDF)
11 Optimistic VC Inequality (PDF)
12 VC Subgraph Classes of Functions; Packing and Covering Numbers (PDF)
13 Covering Numbers of the VC Subgraph Classes (PDF)
14 Kolmogorov's Chaining Method; Dudley's Entropy Integral (PDF)
15 More Symmetrization; Generalized VC Inequality (PDF)
16 Consequences of the Generalized VC Inequality (PDF)
17 Covering Numbers of the Convex Hull (PDF)
18-20 Bounds on the Generalization Error of Voting Classifiers (PDF 1) (PDF 2) (PDF 3)
21 Bounds in Terms of Sparsity (PDF)
22 Martingale-difference Inequalities (PDF)
23 Empirical and Rademacher Processes; Comparison Inequality for Rademacher Processes (PDF)
24-25 Generalization Bounds for Neural Networks (PDF 1) (PDF 2)
26 Talagrand's Convex-hull Distance Inequality (PDF)
27 Consequences of Talagrand's Convex-hull Distance Inequality (PDF)
28 Talagrand's Concentration Inequality for Empirical Processes (PDF)
29 Talagrand's Two-point Inequality (PDF)
30 Talagrand's Concentration Inequality for Empirical Processes (cont.) (PDF)
31 Applications of Talagrand's Concentration Inequality (PDF)
32 Applications of Talagrand's Convex-hull Distance Inequality; Bin Packing (PDF)
33 Generalization Bounds for Kernel Methods (PDF)
34 Optimistic VC Inequality for Random Classes of Sets (PDF)
35 Applications of Random VC Inequality to Voting Algorithms and SVM (PDF)