Skip to content

Details

Week 8 of 8. The session that separates modeling from storytelling.
A model with enough free parameters will fit anything, including noise. Worse, a model can fit your data beautifully and still be unidentifiable — meaning the parameter values you recovered are meaningless. Almost nobody checks. We're going to check, and then we'll look at what members have built over the last seven weeks.
Agenda:

  1. Intros and recap of the series (10 min)
  2. Parameter recovery: simulate from known parameters, refit, see whether you get them back. The single most useful hour in modeling (25 min)
  3. Model comparison: AIC, BIC, and why cross-validation is usually the honest answer (20 min)
  4. Overfitting demonstrated live, with a model we all agree is wrong (15 min)
  5. Show and tell — bring anything you built or read during the series, however rough (30 min)
  6. What we do next season (10 min)
    Bring: a laptop if you want to run the recovery exercise. Otherwise just bring what you made.
    This is an online event. The video link appears on this page once you RSVP.

Related topics

Cognitive Computing
Cognitive Neuroscience
Cognitive Science
Object Oriented Programming

You may also like