MathOptAI.jl
MathOptAI.jl copied to clipboard
lanl-ansi
Embed trained machine learning predictors into JuMP models
People want skip connections, e.g., for ICNN. There's nothing really blocking this. We just need to design the right approach.
Closes #214 Following up from our discussion at JuMP-dev. I have some questions, mainly around how this would be used. Would you want to use this an explicit layer, or...
It would be nice to optionally set the weights/biases as decision variables where the start values are based on the values read in with the network. My use case is...
