Class Description

The main goals of the class are to (1) obtain an overview of current state of the art in the field, (2) carry out a substantial research project, and (3) get practice in research skills such as conducting a literature study, putting together a research and a conference talk, and writing a research paper. The course is organized so that selecting and completing a research project should be as easy as possible.

The first part of the course is an introduction to neural networks. We will first review the most important artificial neural network architectures and algorithms such as backpropagation, deep learning, self-organizing maps, reinforcement learning, and neuroevolution. Information processing in biological neural networks will be reviewed and distributed representations will be introduced as a foundation for connectionist artificial intelligence. Understanding of this material will be tested in an exam. In addition, homework assignments will give students hands-on experience in building simple network models.

The second part is research oriented. Each team of 2 students will select an advanced topic in neural networks, study the literature in depth and give a 50 min presentation to the class on that topic. The presenters will also select a paper on that topic for the rest of the class; each student will prepare two intelligent questions based on the paper.

Based on the specialization, each student (or a team) will then carry out an independent research project. Projects can vary from theoretical analysis of algorithms and development of new mechanisms to applications in engineering and artificial intelligence. The project is presented (in about a 15 min talk) in the class, and documented in a paper.


risto@cs.utexas.edu
Last modified: Thu Aug 15 19:56:21 CDT 2024