CPS 902 Selected Topics in Recognition by Machine

Theme for spring, 2001: Mental Development and Online Learning


Autonomous development of mental skills by machines is a new research field that has drawn increasing attention in the areas of pattern recognition, robotics and artificial intelligence.  Since this new field addresses the development of intelligence through computational modeling, it is expected to be useful for understanding human intelligence.  How much role does mental development play in the generation of human intelligence?  Recent advances in neuroscience have cast serious doubt on the notion that the human genes specify mostly how certain areas of the brain process specific tasks, such as vision, audition and touch. The developmental program in the human genes seems more of a general-purpose nature than what many researchers have thought, and it enables a human to develop its mind through physical experience from infancy to adulthood. In this course, we will study computational models for developing cognitive and behavioral skills autonomously through online, real time interactions with the environment, motivated by mental development in humans and higher animals.

This course is designed for students who are interested in ways human mind develops and how to enable machines to develop their mental skills.  The course is designed in such a way so that graduate students with diverse backgrounds can all take, such as neuroscience, cognitive science, engineering and philosophy.  The teaching material consists of two parts, a draft of monograph written by the instructor for a diverse readership and a collection of papers on the related subjects.

  1. Muddiness of tasks
  2. Overview of AI approaches --- knowledge-based, learning-based, behavior-based, evolutional and the new developmental approach.
  3. Human mental development, results from neuroscience and developmental psychology
  4. Overview of animal learning theories and models
  5. Supervised, reinforcement and communicative learning (how to enable machines acquire language and learn through language)
  6. Architectures for automatic mental development
  7. Automatic generation of representation from data
  8. High-dimensional classification and regression
  9. Sensory system representation and its development (automatic development of well-known filters, such as Laplacian of filters, Gabor filters and wavelets).
  10. Cognitive system representation and its development (cerebral cortex and hierarchical discriminant regression)
  11. Motor system representation and its development (autonomous synthesis of new behaviors)
  12. Attention selection as internal behavior under development
  13. Unification and integration of mental capabilities through development, including vision, audition, touch, language, reasoning, decision making, planning, object manipulation and navigation (information fusion and sensor fusion)
  14. Machine thinking and its development
  15. Consciousness from mental development
  16. Example of experimental artificial developmental systems
  17. Applications, impacts and the future directions



    For more information about this subject, visit website of NSF/DARPA Workshop on Development and Learning

Instructor: John Weng
Office: 2325 Engineering Building; phone: 353-4388; e-mail: weng@cse.msu.edu
Class: 4:10pm - 5:30pm, Tuesdays and Thursdays, room 2400EB.
Text:  Instructor prepared tutorial and articles
Prerequisites: General knowledge about computers and college methematics

Course arrangement:

Each student will work on a related project, either literature survey or a hand-on project, some paper presentations and one project presentation.  No examination.  3 credits.

Back To Weng's Home Page: http://web.cse.msu.edu/~weng/