[SciPy-User] Optimizing odeint without a for loop

[Barrett B]

Rob Clewley <rob.clewley <at> gmail.com> writes:

> 
> On Wed, Jun 4, 2014 at 4:41 PM, Sturla Molden <sturla.molden <at>
gmail.com> wrote:
> > If you are solving Hodkin-Huxley equations you might consider NEURON.
> 
> While I don't disagree that you should consider all the available
> tools (and there is a python interface for it these days), you want to
> balance the overheads of model set up and learning curve with what
> kinds of study you are doing. For instance, for relatively small
> models (a handful of equations, maybe just single compartment
> neurons), even if they are conductance-based, NEURON is overkill, IMO.
> NEURON is particularly well suited for straight-up simulations of
> multi-compartment, highly anatomically realistic neurons. However, if
> you expect to do bifurcation analysis or some exploratory mathematical
> simplifications involving reduced model components, then NEURON will
> not be an appropriate tool. Also, as you mentioned, Barrett, Brian has
> some limitations regarding its numerical schemes, and focuses on
> spiking (I&F-like) models, although NEURON's numerical solvers are
> sound. If you want robust numerics and analytical capabilities,
> PyDSTool is a good place to go (it uses industry standard implicit
> solvers, Dopri and Radau as well as arbitrarily accurate zero-crossing
> detection). So it all depends what you are trying to achieve in your
> project. If you do want more advice, I'd suggest you describe your
> project more.
> 
> -Rob
> 

Right now it's a pretty simple objective: Given a neuronal network, where
each neuron has only a few first-order differential equations, I want to
simulate the spiking or bursting behavior over time.

I took a look at NEURON, and yes, while it seems extremely powerful, it
seems to focus on individual parts of the neuron and not the neurons as
wholes. That would require a lot of computational power, not to mention
setup time.

I'll take a look at PyDSTool.