policy class design

New policy class enforces hard constraints in contextual optimization

Researchers have introduced Legendre-regularized policies, a novel approach to contextual optimization that enforces hard feasibility constraints while maintaining smoothness for gradient-based training. This method par…