Multiple Run Fitter

Using the Multiple Run Fitter, praxis, etc..
Post Reply
magpie
Posts: 23
Joined: Mon Jan 29, 2007 4:13 pm

Multiple Run Fitter

Post by magpie » Tue Jul 10, 2007 3:14 pm

I am trying to fit the Recovery Cycle data from the existing model
with the experimental data. However, since the Recovery cycle data are
obtained from analyzing the results of many runs, rather than just 1
run as in the tutorial. Specically, my hoc file does:
- Run once, determine if there is an action potential
- continue with binary search until find out the threshold current
- change the inter-stimulus delay and perform the threshold tracking again
- after each loop, the threshold and delay are save into a matrix
called Fitdata
Hence, I would like set up the Fitter to run the whole process of
Recovery Cycle each time it generates an Error Value. However, I have
2 difficulties:
- How to use the matrix Fitdata as the Fitness Variable since it
does not appear in the list
- How to set up the Fitter to run the whole process of Recovery
Cycle (the procedure is contained in the proc "recov()")

Thanks

ted
Site Admin
Posts: 5623
Joined: Wed May 18, 2005 4:50 pm
Location: Yale University School of Medicine
Contact:

Post by ted » Mon Aug 06, 2007 11:18 am

Approach the task as a "function optimization" problem.

1. Define a func thresh() that takes a single argument (ISI) and returns the
corresponding spike threshold.

2. Wrap this inside another func that sets model params from conveniently named
proxy variables. For example, if the model params you want to tune were soma cm
and axon Ra,

Code: Select all

CM = 1 // whatever appropriate initial value you like
RA = 125
func parameterized_thresh() {
  soma cm = CM
  axon Ra = RA
  return thresh()
}
At this step you're at a point comparable to "Step 1" of the function fitter part of the MRF
tutorial http://www.neuron.yale.edu/neuron/stati ... rtmrf.html

3. Start an MRF, add a Function Fitness generator, and paste your experimentally
observed time course of spike threshold into it (your data are in the form
isi0 thresh0
isi1 thresh1
etc.
right?)

4. Specify the name of the function that is to be compared against the experimental data,
i.e. use
Fitness / Expression with $1
and in the edit field enter
parameterized_thresh($1)

5. Use
Parameters / Add Parameter
to specify the parameters that are to be adjusted. In this example, they are CM and RA.

Now you're ready to test the Generator etc., as described in the tutorial.

Post Reply