Project F: Identifying Neurons by Firing Features
Investigator: Marc Sommer, Dept. of Neuroscience, U. Pitt
Team: Bellone Gaia, Shuhei Okomura

=========================================================

Mon Jan 22 10:11:24 2007

Met Marc in December

Marc has a pretty preset agenda for them so far:

1. Model fitting and selection

   Poisson spike model with 1 or 2 parameters

   More general point process models

`  NEEDS: 
     - how to fit these model
     - are there R packages to do this?

2. Classification based on best-fitting model

   2 major categories of neurons
     - Exitatory
     - Inhibitory

   Several subcategories but they're not sure 
   how many of these Marc already knows about

DATA 

   Getting data at next meeting (this week?) with marc.

   Expect 56 neurons x 3 derived features

   Is there a richer data set behing this?

QUESTIONS:

   From marc's presentation I was expecting feature selection and
   unsupervised clustering.  But Marc has a more focused, parametric
   model based approach set up.  Is this because of the small data
   set?

====================================================================