🚀 go-pugleaf

Paul Bush (paul@helmholtz.sdsc.edu) wrote:


: Why are you writing your own simulator rather than using, for example,
: NEURON, written by Michael Hines? This simulator has been refined over
: several years and solves many of the problems encountered when writing
: your own. It uses very fast and efficient numerical integration routines;
: My model of 100 9-compartment cortical neurons connected at a density of
: 10% simulates 500 ms in 9 minutes of CPU time on a MIPS 3000. Its
: modular nature allows the easy addition of arbitrary processes such
: as the graded/electical synapses you mention. It has an Interviews front
: end for graphical display, control and online modification of the
: simulation.

: To produce a simulator as good as NEURON you would need to spend
: several years programming and testing - that time would be better spent
: doing the actual simulations. In addition users of NEURON share their
: simulation code, giving you access to many types of ion channels, synapses
: etc.

: You can ftp NEURON from neuron.neuro.duke.edu.

: Paul

Why my own model? Three reasons, really. First, have you ever tried to
port NEURON to a Mac (my initial development machine)? 8( Second, I needed
to roll my own synaptic models, and it was easier if I didn't have to
learn a model at the same time. Third, I regarded putting together my own
model before using someone else's to be a very valuable learning
experience. I'll probably use GENESIS from here on out, but doing my own
taught me a lot about why explicit codes don't work well and about how to
adapt the existing tools to my own needs. I recommend this path (building
a throw-away model or two first) to any graduate student who is serious
about modeling the neural system. 

--
Harry Erwin
Internet: herwin@gmu.edu