Course Description:

This hands-on exploratory course introduces students to the principles of Neural Networks. It starts with a model of a single neuron that is then used as a basic information processing unit in a number of experiments in several network topologies. The focus is on conceptual modeling rather than on theoretical foundations.

Course Outline

  • Interactive Activation and Competition
  • Constraint Satisfaction in Parallel Distributed Models (PDP)
  • Learning in PDP Models: The Pattern Associator
  • Training Hidden Units with Back Propagation
  • Competitive Learning
  • Simple Recurrent Network
  • Recurrent Backpropagation
  • Temporal-Difference Learning

Student Outcomes:

After graduation from the course, the students will be able to:

  • organize and express ideas concerning the foundations of neural networks clearly and convincingly in oral and written forms,
  • recognize applicability of methods based on neural networks to real-world problems,
  • identify appropriate neural networks to solve specific real-life problems,
  • formulate real-life problems in terms suitable for processing with neural networks,
  • implement neural network-based solutions to real-life problems,
  • evaluate neural network-based solutions to real-life problems through experimentation that includes data collection and analysis,
  • further their knowledge of the field by applying the foundations studied in this course.

Instructor: AJ
Bieszczad

Office: Sierra Hall Room 3315
Email: aj.bieszczad@csuci.edu
Phone: (805) 437-2773
Office hours: Wednesdays 2:00 – 4:00 pm

Instructor Communication Policy:

I will respond to your inquiry within 24 hours Mon-Fri. If I do not reply in this timeframe, please assume I did not receive your email and contact me again.

Please manage the communication diligently.

Textbook:

Explorations in Parallel Distributed Processing with MATLAB
(http://www.stanford.edu/group/pdplab/pdphandbook/)

Supporting book:

Parallel Distributed Processing, David E. Rumelhart,
James L. McClelland
Publisher: The MIT Press (July 29, 1987)
ISBN-13: 978-0262631129
(NOTE: A
PDF version available in Blackboard and – in pieces – from http://www.stanford.edu/group/pdplab/resources.html
)

Instructional Approach:

The class will be managed completely asynchronously online. Students must use CI Learn to access all course material, submit assignments, and check grades. A discussion board is available for asking questions.

The course will be a sequence of directed self-studies. Every two weeks, a new topic to study over the coming course segment will be announced. Students will study theoretical foundations and then experiment with the corresponding neural network model. A submission link for a given assignment will appear at the  beginning of each segment of the course, and will disappear after the deadline for that week.

All deadlines are in the course calendar.

Please note that
late submissions will absolutely not be allowed

Tools:

CS Students are requested to use Python 3 with NumPy, SciPy, and Matplotlib to implement the neural netowkr models and use them for experiments described in the assignments. No GUI is necessary; just a research note with the experiment results supported by visuals plotted with Matplotlib.

Math students may use MATLAB-based tools that are freely available from Stanford University at http://www.stanford.edu/group/pdplab/resources.html.

Grading:

The final grade in the course will be based completely on the quality of the submissions to the assignments; there will be no exams in this course. There will be a number of assignments with multiple exercises. Each assignment carries the same weight.

Course Credits: 3 units
Grade Letter
> 96% – A+
> 88% – A
> 80% – A-
> 75% – B+
> 68% – B
> 62% – B-
> 55% – C+
> 48% – C
> 41% – C-
> 34% – D+
> 27% – D
> 20% – D-
< 20% – F

All grades are computed automatically by the grading system in Blackboard from points allocated by the instructor. The grades might be adjusted at the instructor’s discretion.

Please make a note that the grade scale will be from 0% to 100% rather than more common 60%-100%.

Academic Honesty

The university, the course, the labs and the instructor are here for the students so they can acquire sufficient knowledge to open a window of opportunity for them in their future careers. Any academic dishonesty limits the student’s chances to succeed.

Please consult the Academic Catalog for the details on the CSUCI’s academic code of honor.

Students with
Disabilities

Students with disabilities needing accommodation should make requests to the CSUCI Disability Resource Programs.
Please discuss approved accommodations with the instructor.

Subject to Change

This syllabus and schedule are subject to change in the event of extenuating circumstances.