Oscar Dowson
Oscar Dowson
I'm going to close this issue because I don't have plans to implement it. A pretty fundamental design decision of SDDP is that we purposefully don't do well "off-chain". The...
Ah I knew there was something! A student of Dave is trying to do something similar.
💯 This was the next step I had in mind: They're coming at it from a different direction/motivation, but it's the same core technique.
Yeah I thought of this. I initially considered looping through the cuts evaluating them before I realized I could reach in and find the right `JuMP.value(theta)`. I guess the point...
There are still some issues with this, but it's a useful start
I have not implemented it. All code for SDDP.jl is contained in this repo.
Hi @FSchmidtDIW, how'd you get on? Here's a place that might be useful to start https://github.com/odow/SDDP.jl/blob/3b2570b78c38c9e2073538a7b14b32b04fa12fe2/src/plugins/forward_passes.jl#L323-L372
I strongly dislike any methods which require modifying existing cut intercepts. It is not something that is easily achievable in SDDP.jl. I don't know if you should spend time working...
If you're thinking about this: * The forward pass can be computed independently of the current policy (the cuts) * You don't even need the sequence of states in the...
> Working on capacity expansion models, it appears useful for the first stage at least but I suppose the existing RegularizedForwardPass already does that Yes, now thiiiiiis makes sense. Feel...