Week | Name | Summary |
---|---|---|
1 | Basic course information | |
2007-07-27 Lecture | Why you should take CS705 | |
2 | 2007-07-31 Lecture | Bayesian learning, smoothing and conjugate priors, from least-square to ridge regression |
2007-08-03 Lecture | Ridge and Lasso regression, non-linear regression, classification vs. regression | |
2007-08-04 Lecture | Bayes-optimal and conditional probabilistic classification, discriminants for multivariate Gaussian distributions, loss functions for regression and classification | |
3 | 2007-08-07 Lecture | Scilab Tutorial by Dhaval |
2007-08-11 Homework 1 | For coding and submission details please contact the TAs. | |
4 | 2007-08-14 Lecture | Singular value decomposition by Professor Ranade |
2007-08-17 Lecture | Singular value decomposition, linear least square fitting, principal components analysis, eigenvalues and eigenvectors, the Perceptron algorithm | |
5 | 2007-08-21 Lecture | Linear least square fitting, pseudoinverse, iterative methods, Fisher's linear discriminant |
2007-08-24 Lecture | Fisher's linear discriminant, logistic regression, maximum entropy classification | |
2007-08-25 Homework 2 | There may be minor edits and fixes but the broad questions will remain unchanged. I am posting a preliminary version so that students can start early and complete before midterm exams. | |
6 | 2007-08-28 Lecture | Connections between SVM and Logistic Regression, duality theory recap |
2007-08-31 Lecture | Dual SVM QP, the kernel trick | |
7 | 2007-09-04 Lecture | Introduction to StructSVM (Prof. Sunita) |
2007-09-07 Lecture | TsochantaridisJHA2005structsvm proofs (Prof. Sunita) | |
2007-09-09 Midterm exam | With some solutions and grading policies. | |
9 | 2007-09-18 Lecture | Midterm review, StructSVM revisit, ramifications |
2007-09-21 Lecture | Linear-time linear SVM, SVM for non-decomposable loss functions | |
10 | 2007-09-25 Lecture | Ordinal regression and ranking |
2007-09-27 Lecture (8:30--9:30am Thu) | Designing kernels for regression and classification, RKHS basics | |
2007-09-28 Lecture (11:30am--12:30pm Fri) | More about kernels, infinite dimensional spaces, eigen functions, RKHS, representer theorem | |
11 | 2007-10-05 Lecture | RKHS and representer theorem, variations on SVM optimization |
12 | 2007-10-09 Lecture | CCCP review; transductive learning |
2007-10-12 Lecture | Probabilistic model learning, expectation maximization | |
13 | 2007-10-16 Lecture | Multinomial mixture, EM, optimization, variational approach |
2007-10-19 Lecture | NMF, dyadic aspect model, cross association and coclustering | |
14 | 2007-10-23 Lecture | Ensemble learners, bagging and boosting |
2007-10-26 Lecture | More about bagging and boosting, rate distortion theory | |
Homework 3 | ||
15 | 2007-10-30 Lecture | Information bottleneck, approximating (functions of) distributions by Monte-Carlo sampling |
2007-11-03 Lecture (Sat, not Fri) | Metropolis-Hastings; Generalization theory | |
16 | 2007-11-06 Lecture | Generalization theory and risk bounds |
17 | 2007-11-13 Lecture | Empirical processes, risk bounds, growth function, VC theory |
2007-11-16 Lecture | Risk bounds, symmetrization, stability-based generalization bounds | |
2007-11-18 Lecture | Course wrap-up | |
18 | Final exam | 9:30am--1pm, Friday 30th November, 2007 |