Course - Prerequisites - Textbooks - Grading
Course Goals and Description
This is a graduate level course on neural networks. The course covers the following topics:
|
Historical introduction. Connections to biological modeling, computational neuroscience, and cognitive modeling. Artificial/Natural NNs.
|
|
Basic concepts of feedforward networks. Single layer perceptrons. Multi-layer perceptrons Activation functions. Higher-order NNs. Error functions. Regression. Classification.
|
|
Universal approximation properties.
|
|
Bayesian probabilistic framework and Bayesian statistical approach to NNs. Maximum likelihood approaches.
|
|
Learning. Gradient Descent. Backpropagation.
|
|
Recurrent NNs. Learning in recurrent NNs.
|
|
NN differential equations.Dynamical systems.
|
|
Hopfield model. Bolotzmann machines. Optimization.
|
|
Unsupervised learning. K-means. Self-organizing maps.
|
|
Bias-variance tradeoffs.
|
|
Weight sharing. Cross validation methods. Weight decay. Ensemble methods.
|
NNs will be applied to a variety of data sets including time series and spatial data in diverse fields ranging from biology to finance.
The official catalog description is:
ICS 276A: Introduction to concepts of artificial neural networks (ANNs). Architectures for supervised and unsupervised networks. Mathematics of learning and performance rules.
Course - Prerequisites - Textbooks - Grading
Prerequisites
A basic understanding of discrete and continuous mathematics, calculus, and probability theory, as well as proficiency in programming, or consent of instructor.
Course - Prerequisites - Textbooks - Grading
Textbooks
Neural Networks for Pattern Recognition by Christopher M. Bishop (Oxford University Press)
Bioinformatics: the Machine Learning Approach by Pierre Baldi and Soren Brunak (MIT Press)
Course - Prerequisites - Textbooks - Grading
Grading
Students will read articles from the literature. Grading will be based on participation in class discussions, one exam, and a project with final presentation resulting in a brief (5--10 pages) conference-style written report. Additional assignments can include homeworks.
|