NUIM CS401 F2006
Mon 12:00-12:50 CS2
Tue 15:00-15:50 JohnHume6
Instructor: Barak A.
Pearlmutter, barak@cs.nuim.ie
Office: Hamilton Institute (NUIM,
Rye Hall, South Wing, room 5)
Office hours: you are welcome any time, just drop on by. (Afternoons
are best, excepting Fridays.) Or feel free to email or ring me up
(x6394) and make an appointment. If people would prefer that I set
aside some particular weekly times, let me know and I will do so.
Text
We will use notes made available on the web.
Lectures
- (18-Sep-2005)
Introduction to Machine Learning, pre-requisites
- (19-Sep-2005)
Introduction to R
(see NOTES.R)
- (25-Sep-2005)
Misc Definitions, and
Intro to Linear Classifiers, ie Perceptron
(notes)
- (26-Sep-2005)
The Perceptron Learning Rule.
The "homogeneous coordinates" trick
(notes)
- (2-Oct-2005)
Linear regression, gradient descent
- (3-Oct-2005)
More gradient descent
Backpropagation (part 1 of 2)
- (9-Oct-2005)
Backpropagation or reverse-mode AD (part 2 of 2)
- (10-Oct-2005)
Convergence of vanilla gradient descent of a linear unit with
quadratic error
- (16-Oct-2005)
Generalisation Curves
- (17-Oct-2005)
Example of unsupervised learning: k-means clustering
- (23-Oct-2005)
Maximum Likelihood Estimation: definition and toy examples
(coin, Gaussian)
- (24-Oct-2005)
EM of a simple Gaussian mixture model
- (6-Nov-2005)
Hidden Markov Models or HMMs (1/3)
- (7-Nov-2005)
Hidden Markov Models or HMMs (2/3)
- (13-Nov-2005)
Hidden Markov Models or HMMs (3/3)
- (14-Nov-2005)
graphical models: intuitions (1/2)
- (20-Nov-2005)
graphical models: Energy, Boltzmann distributions, Monte-Carlo (2/2)
- (21-Nov-2005)
Support Vector Machines I:
- dual representation of weights
- maximum margin hyperplane
- (27-Nov-2005)
Support Vector Machines II:
- quadratic programming
- the kernel trick!
- (4-Dec-2005)
Multiplicative updates
- (5-Dec-2005)
Boosting
(notes)
- (11-Dec-2005)
Reinforcement Learning. Policy-value iteration, Q-learning, TD
(notes)
- (12-Dec-2005)
Case Studies: ALVINN, PAPnet, TDNN for E-set,
handwritten digit recognition, TDgammon
Useful Materials
Author's lecture notes
from Machine Learning by Tom Mitchell.
Code written in class, usually scrubbed up a bit.
Other Notes