|
週次 |
日期 |
單元主題 |
|
第1週 |
9/16,9/18 |
1. Introduction
2. Supervised learning |
|
第2週 |
9/23,9/25 |
1. Review Fisher linear discriminant analysis, linear SVM.
2. Matlab tutorial on linear SVM |
|
第3週 |
9/30,10/02 |
1. Finish Chapter 2 (lecture notes revised).
2. Starting Chapter 3 on Friday by Prof. Tu |
|
第4週 |
10/07,10/09 |
continue on Chapter 3 |
|
第5週 |
10/14,10/16 |
1. examples and matlab demo.
2. Kernel method for nonlinear SVM |
|
第6週 |
10/21,10/23 |
1. Chapter 5.
2. Partial least squares. Slides on 討論看板. |
|
第7-1週 |
10/28 |
1. PLS matlab demo.
2. continue on nonlinear SVM.
3. discuss your final project topic selection.
|
|
第7-2週 |
10/30 |
1. SSVM -a smooth SVM algorithm implemented in the primal space. 2. Two auxiliary techniques (reduced kernel and uniform design) for fast computation. |
|
第8-1週 |
11/04 |
reduced kernel and matlab demo |
|
第8-2週 |
11/06 |
1. Lagrangian SVM.
2. Chapter 6, start with kernel PCA. |
|
第9-1週 |
11/11 |
11/11 (Wed), no class, take-home midterm due. |
|
第9-2週 |
11/13 |
Midterm problem 1 presentation and discussion. You will be asked to present your problem 1. No need to make extra preparation, just use your submitted paper. |
|
第10-1週 |
11/18 |
Lagrangian SVM |
|
第10-2週 |
11/20 |
1. midterm paper revision due.
2. dimension reduction by kernel SIR. |
|
第11-1週 |
11/25 |
2 projections commonly seen in dimension reduction: orthogonal projection, oblique projection. |
|
第11-2週 |
11/27 |
1. support vector regression. 2. KSIR revisited. |
|
第12-1週 |
12/02 |
SVR and KSIR matlab demo. |
|
第12-2週 |
12/04 |
Review: basic kernel theory, kernel trick, SVM & SVR in primal and in dual, auxiliary techniques and tools.
Final project preview: boosting, nonlinear dr (locally linear embedding, Laplacian LLE, Hessian LLE), Lasso and its dual, classes of kernels, anti-correlation kernels, locality preserving projections. |
|
第13-1週 |
12/09 |
Homework-4 (on SVR and KSIR) due, this is the last homework. |
|
第13-2週 |
12/11 |
Final project oral presentation. Boosting (李尚文). Nonlinear dimension reduction -LLE (陳帥,吳佩勳). |
|
第14-1週 |
12/16 |
Oral presentation. Locality preserving projections (宮嵊益). |
|
第14-2週 |
12/18 |
Student oral, Lasso and its dual (郭晏伶,湯泉發). Anti-correlation kernels (門諦風). |
|
第15-1週 |
12/23 |
Mini-workshop on Friday, no class today. |
|
第15-2週 |
12/25 |
TIMS Seminar in Statistical Methodology. 9:30-12:00, at New Math 308. Attendance is mandatory and credited. |
|
第16-1週 |
12/30 |
Classes of kernels (陳幼剛). Nonlinear dimension reduction -Hessian local linear embedding (何昊). |
|
第16-2週 |
01/01 |
Holiday, no class. |
|
第17-1週 |
01/06 |
1. Hessian LLE.
2. Homework 4. |
|
第17-2週 |
1/08 |
1. Final project paper due. All the students slides can be accessed at 作業區-作業觀摩.
2. Tutorial and matlab demo on Neural Networks -Basic. |