About Constrainted Deep modeling

[xtzd]

Hi, I am new to this repo. I wonder if the PyEPO could solve such work:

a simple multi-demensional regression supervised deep learning problem.

suppose x : (x1,x2,x3,...xm)
suppose y : (y1,y2,y3,...yn)

y_pred = NeuralNetwork(x)
and we want to minimize

loss(y_pred,y_true)

under a linear inequality constraint R ,such as

ay1+by2+...n*yn<A

We want to train the Network.

Can PyEPO solve such problem? If it could, could you provide simple paradigm?
Thank you!

[LucasBoTang]

Hi xztd,

PyEPO is not suitable for your problem. It is designed to deal with optimization problems with unknown cost coefficients. Here you want to predict a solution with a linear inequality constraint directly.