Oscar Dowson
Oscar Dowson
@adow031 #423 adds a warning and changes the headers. But https://odow.github.io/SDDP.jl/stable/guides/improve_computational_performance/#Initialization-hooks is still the recommended way to do this. We need to run code on each process _after_ the model...
Here you go: You still need to define an optimizer for the master process. The async callback is just to get new pointers on the remote processes. Please suggest improvements...
> when I used settings for Gurobi (i.e. optimizer_with_attributes) it didn't crash, but SDDP only used process 1. Can you elaborate?
> It only reports cuts from process 1. Why are they workers 6-9? Did you load them in an existing session? What happens if you start a new one? You...
Very weird. I'm not aware of why this would happen... Do you have the call to `train` you used?
Ideally, one of @pulsipher's examples from InfiniteOpt: https://github.com/pulsipher/InfiniteOpt.jl/tree/master/docs/src/examples
If anyone reading this has nice examples of an optimal control type problem they want to share, please reach out.
Thanks for the pointer! Here's the SDDP.jl equivalent ```Julia using SDDP import Distributions import Gurobi import Plots function main( ρ = 0.020, # discount rate k = 90.0, # utility...
> The theory also says that if households are credit constrained, B >= 0 (as you do), then consumption will be less smooth. But in this example we are not...
> Can you resolve the problem w/o B>=0? It will be the same answer because the wealth constraint is never binding. > I think this problem actually has a closed...