Sam Isaacson

@yewalenikhil65 Yes, one could create a particle-tracking version of spatial-stochastic models, i.e. a Brownian Dynamics model. That could potentially be built on top of Agents.jl (though there might be more...

In terms of discrete, probably we should just target unstructured grids (i.e. graphs) first. Cartesian grids are really just a special case and could be added subsequently as an optimization.

Yeah, as long as it isn't breaking this would be nice to add.

I think we need a bit of thought about how we want to handle this (i.e. maybe we should be storing all the network properties within some object and just...

I’m not sure what you mean? Explaining what the various things you are calculating are, and their significance, as part of the tutorial would make sense.

No rush! I hope you feel better soon. I think the stuff on model reduction should be skipped for now. In particular, that method doesn't seem to work that well...

Sounds great!

Hi, unfortunately the GPU parallelization refers to the ODE and SDE solvers. I'm not aware that there are any real good (exact) GPU-based Gillespie algorithms, but I haven't kept up...

We should indeed add a tutorial that shows how to use GPUs for ODE/SDEs. @ChrisRackauckas would that make sense to have in the code-optimization tutorial you were planning to work...

Which Gillespie method algorithm did you use on GPU? How big was your reaction model? The Gillespie methods we expose are all in JumpProcesses.jl, but I doubt any of them...