Joe Lau

Seminars in cognitive science

This page covers only the second part of the course. The first half of the course will be taught by Dr. Hayes.

Announcements

  • [8 May] Here are some details about the final report.
  • [10 Apr] I just sent out an email about the details of the presentation session to the cogn3002 mailing list. Please check.
  • [27 march] Before noon 10 Apr, you should email me details about your project: (a) if you work as a team, your team members and their full names, (b) the topic of your project and some details about your progress and approach. Of course, if you encounter any difficulties and if there is anything you do not understand, please feel free to send me an email or making an appointment.

Arrangements

  • Dr. Lau will give one introductory lecture on Feb 12. You will be studying the required course text on your own afterwards. You are required to study chapters 1-8 of the book on your own and complete all the exercises.
  • You are also required to design your own simulation project using the tlearn package and give a 10 minute presentation on your project near the end of term and submit a report.
  • You can either work on your own project or work as a team (max 3 persons per team). If you work as a team, your team has to agree and let me know the relative contribution of each member (e.g. 40%/60% for a 2-membered team).
  • If there are fewer than 9 presentations, they will all be held on 7 May. Otherwise some of them will be held on April 30.
  • You should email your reports to me at before 11pm 14 May 2007.
  • Assessment: presentation (30% of total marks) + report (20%) +/- participation.

Course material and readings

  • [Required text] Kim Plunkett, Jeffrey L. Elman. 1997. Exercises in Rethinking Innateness: A Handbook for Connectionist Simulations. MIT Press. isbn:0262661055 It would be best for you to get a copy of this book. You can ask the university bookstore to order a copy for you. HKU also has an ebook version online.
  • [Required simulation software] tlearn: http://www.columbia.edu/cu/psychology/courses/3205/tlearn/ Unfortunately the software contains quite a few bugs and crashes rather frequently. So save your work regularly.
  • Other useful books
    • Reilly and Munakata (2000). Computational Explorations in Cognitive Neuroscience. MIT Press.
    • McLeod, Plunkett, Rolls (1998). Introduction to Connectionist Modelling of Cognitive Processes. OUP.
  • Main.ConnectionismIntroduction

Issues to bear in mind

  • Why use different activation functions? How would the different functions affect learning?
  • What does the bias unit do?
  • How is momentum different from learning rate?
  • What is the perceptron learning rule?
  • What is linear separability, error function?
  • What is the purpose of cluster analysis?