 |
Description
Neural networks provide a model of computation drastically different
from traditional computers. Typically, neural networks are not explicitly
programmed to perform a given task; rather, they learn to do the task
from examples of desired I/O behavior. The networks automatically generalize
their processing knowledge into previously unseen situations, and they
perform well even when the input is noisy, incomplete or inaccurate. These
properties are well-suited for modeling tasks in ill-structured domains
such as speech recognition, motor control and cognitive processing. Artificial
neural network models are inspired by biological neural networks. The
course begins with an overview of information processing principles in
biological systems and the organization of the human brain. The core of
the course consists of the theory and properties of major neural network
algorithms and architectures. The students will have a chance to implement
and try out several of these models on practical problems. By the end
of the course, the student will be able to assess the applicability of
neural networks for a given task, select an appropriate neural network
paradigm, and build (i.e. configure and train) a working neural network
model for the task.
|
|
|
Prerequisites
Prerequisite: the following courses, with a grade of at least "C" in
each: CS 307, CS 310, CS 315, CS 328, CS 336, M 408D, and PHL 313K.
|
|
