UC Berkeley
Department of Electrical Engineering and Computer Science
Department of Statistics
EECS 281A / STAT 241A Statistical Learning Theory
Solutions to Problem Set 3
Fall 2011
Issued: Monday, October 10, 2011
Due: Monday, October 24, 2011
Reading: F
UC Berkeley
Department of Electrical Engineering and Computer Science
Department of Statistics
EECS 281A / STAT 241A Statistical Learning Theory
Problem Set 5
Fall 2011
Issued: Thursday, November 10, 2011
Due: Wednesday, November 30, 2011
Problem 5.1
Conv
UC Berkeley
Department of Electrical Engineering and Computer Science
Department of Statistics
EECS 281A / STAT 241A Statistical Learning Theory
Problem Set 3
Fall 2011
Issued: Monday, October 10, 2011
Due: Monday, October 24, 2011
Reading: For this probl
UC Berkeley
Department of Electrical Engineering and Computer Science
Department of Statistics
EECS 281A / STAT 241A Statistical Learning Theory
Problem Set 1
Fall 2011
Issued: Thurs, September 8, 2011
Due: Monday, September 19, 2011
Reading: For this pro
UC Berkeley
Department of Electrical Engineering and Computer Science
Department of Statistics
EECS 281A / STAT 241A Statistical Learning Theory
Solutions to Problem Set 1
Fall 2011
Issued: Thurs, September 8, 2011
Due: Monday, September 19, 2011
Reading:
CS281B/Stat241B. Statistical Learning Theory.
Lecture 12.
Peter Bartlett
Support vector machines
Hard margin
Detour into optimization
Dual form of SVM: support vectors
Kernels
SVM and the convex hull of the data.
1
Recall: Perceptron convergence the
CS281B/Stat241B. Statistical Learning Theory. Lecture 18.
Nan Ye
17 Mar 2016
In last lecture, we introduced the boosting problem for binary classification, and we have
seen AdaBoost as a greedy variational algorithm to solve it. We showed that weak learni
CS281B/Stat241B. Statistical Learning Theory.
Lecture 13.
Peter Bartlett
Review: Support vector machines
Hard margin SVM, support vectors, kernels
Soft margin support vector machines
Quadratic program
Dual
-SVM
Representer theorem
1
Review: Support
CS281B/Stat241B. Statistical Learning Theory. Lecture 17.
Nan Ye
15 Mar 2016
In this lecture, we introduce AdaBoost, a meta learning algorithm which can be used to
improve the performance of other learning algorithms.
1
The Boosting Problem
Given a traini
CS281B/Stat241B. Statistical Learning Theory.
Lecture 23.
Peter Bartlett
Online convex optimization.
1. Regularized minimization
Bregman divergence
Regularized minimization equivalent to minimizing latest loss
and divergence from previous decision
Con
UC Berkeley
Department of Electrical Engineering and Computer Science
Department of Statistics
EECS 281A / STAT 241A Statistical Learning Theory
Problem Set 2
Fall 2011
Issued: Wednesday, September 28, 2011
Due: Monday, October 10, 2011
Reading: For this
UC Berkeley
Department of Electrical Engineering and Computer Science
Department of Statistics
EECS 281A / STAT 241A Statistical Learning Theory
Solutions to Problem Set 2
Fall 2011
Issued: Wednesday, September 28, 2011
Due: Monday, October 10, 2011
Readi
UC Berkeley
Department of Electrical Engineering and Computer Science
Department of Statistics
EECS 281A / STAT 241A Statistical Learning Theory
Iterative proportional tting and exponential families
Fall 2011
1
Introduction
In previous lectures, we have c
UC Berkeley
Department of Electrical Engineering and Computer Science
Department of Statistics
EECS 281A / STAT 241A Statistical Learning Theory
Problem Set 4
Fall 2011
Issued: Monday, October 24, 2011
Due: Monday, November 7, 2011
Reading: For this probl
UC Berkeley
Department of Electrical Engineering and Computer Science
Department of Statistics
EECS 281A / STAT 241A Statistical Learning Theory
Problem Set 4
Fall 2011
Issued: Monday, October 24, 2011
Due: Monday, November 7, 2011
Reading: For this probl
UC Berkeley
Department of Electrical Engineering and Computer Science
Department of Statistics
EECS 281A / STAT 241A Statistical Learning Theory
Problem Set 5
Fall 2011
Issued: Thursday, November 10, 2011
Due: Wednesday, November 30, 2011
Total: 40 points