Course: 15-880(A) -- Introduction to Neural Networks

Instructors:  Dave Touretzky and Alex Waibel

Date and time:  Tuesdays and Thursdays from 4:30 to 5:50 in Wean Hall 5403,
starting September 17 and running through early December.

Recommended reference:  Introduction to the Theory of Neural Computation,
by Hertz, Krogh, and Palmer.  Addison-Wesley, 1991.  Should be available
in the CMU bookstore.

Registration requirements: prior permission of the instructor is required,
except for CS/RI/ECE/Psych grad students.  Students should have a sound
mathematical background (vector calculus, basic differential equations) and
good computing skills.  No prior experience with neural networks is
necessary.  The course is worth 6.0 units for students who do a project and
take the final exam.

Partial Contents:
  * Basic pattern recognition concepts, including both connectionist and
    non-connectionist approaches.

  * Major types of neural network models:  Hopfield nets, Boltzmann machines,
    multilayer perceptrons, recurrent networks, interactive activation
    models, etc.

  * Learning algorithms associated with these models, including backprop,
    counterprop, the Boltzmann learning procedure, and so on.

  * Applications to speech recognition, autonomous navigation, etc.

  * Connectionist approaches to knowledge representation and natural
    language understanding.

  * Basic neurophysiology:  neurons, synapses, ion channels, ...

  * Neuroanatomy of the visual system.