Zachary Sunberg

Also, could we just fix #9 and just have `solve(solver, pomdp)`? I really don't like the policy argument.

> Actually, I found it pretty useful over the last few days to resume a solving process. > Let say you have ran Value Iteration for some iterations, you can...

Awesome! what were the results? I think it is pretty cool that we designed things so this script can be that concise. Also you say "still needs shared arrays to...

Maybe... I personally think that channels or the other built-in mechanisms would be fine and I would rather not have the extra dependency, but I am not totally opposed to...

Yes, that seems like a good idea - we also could just switch to a Compose-based approach. If it is for the lasertag problem. I have a compose-based visualization for...

Yes, you are correct. For now, as a hack, you can override it yourself by overriding the `initialstate` function, for example: ```julia POMDPs.initialstate(p::TabularPOMDP) = POMDPModelTools.Deterministic(1) ``` Eventually, we should put...

Hi @wbrannon , thanks for reporting the error. POMCPOW is designed for POMDPs, so there is no `solve` method for MDPs. You may be able to get it to work...

Ah, shoot! Well, if you want, we can make the switch to `AbstractArray`s without actually having any concrete type to use. There may be a n-dimensional sparse representation out there...

Only gridworld has them for now, partly because we haven't made a standard visualization interface yet: https://github.com/JuliaPOMDP/POMDPs.jl/issues/107

Yes, this would be great!