Machine Learning

Machine Learning is the study of how to build computer systems that learn from experience. It is a subfield of Artificial Intelligence and intersects with statistics, cognitive science, information theory, and probability theory, among others. The course will explain how to build systems that learn and adapt using real-world applications from industry and science (e.g., learning to classify astronomical objects, to predict medical diagnoses, to play chess, etc.).

Administrative Details

Here you will find administrative information for the Spring 1388.

. Instructor: Dr. Mohsen Afsharchi, afsharchim at znu.ac.ir

. Lectures: 8-11 Saturdays

. Office Hours: Wednesday and Thursday

. Prerequisites: Clear understanding of probability, common data structures, algorithms, standard programming and preferably some preliminary AI concepts.

Textbooks

The required textbooks for this course are:

.

Machine Learning

Tom Mitchell, McGraw Hill, 1997

.

Pattern Recognition and Machine Learning

Christopher M. Bishop, Springer (2006)

A supplementary textbook (recommended, but not required) is:

. Introduction to Machine Learning, by Nils J. Nilsson, Available Online

. Artificial Intelligence: A Modern Approach (Second Edition), by Stuart J. Russell, Peter Norvig, Prentice Hall 2002 .

Lecture material will be drawn from textbooks, as well as from some of the recent online materials.

Course Syllabus and Lecture Notes

Foundations of Machine Learning: Introduction(PDF), A good Introduction By Rob Schapire (PDF)
Concept Learning: General to Specific Ordering (PDF) Courtesy of Tom Mitchell
Decision Trees: Information gain (PDF)
Weka (Open Source Machine Learning Software in Java), (Collection of Datasets)
Instance Based Learning: K-nearest neighbor(PDF)
Curve Fitting: Regression: Simple Concepts (PDF)
Linear Models for Regression: Maximum Likelihood, Least Squares and Bayesian Linear Regression (PDF)
Clustering: K-means (PDF) , Mixtures Models, EM (PDF)
Kernel Methods: Linear Learnable Models, Dual Representation and Feature Space (PDF)

Course Work and Evaluation

HomeWork 1:

You are given five data sets (see below), each containing predictor-values (x-values) and corresponding target-values (t-values). For each data set you should train a linear regression model to predict target-values (t-values) for future prediction-values (x-values), i.e. for each data set, you should select feature function φ, construct model matrix Φ from the prediction values, and estimate weight vector w. From these, you can predict new values as the mean of the predicted distribution, i.e. t=y(x,w)

DataSets

Possibly Interesting URLs

Here is an ad hoc collection of relevant ML links and interesting tidbits. If you know of other good stuff to share with your classmates here, please let me know, and I will try to add it.