A Python package of accelerated first-order algorithms for solving relatively-smooth convex optimization problems
minimize { f(x) + P(x) | x in C }
with a reference function h(x), where C is a closed convex set and
- h(x) is convex and essentially smooth on C;
- f(x) is convex and differentiable, and L-smooth relative to h(x), that is, f(x)-L*h(x) is convex;
- P(x) is convex and closed (lower semi-continuous).
Publication:
https://arxiv.org/abs/1808.03045 (opens in new tab)
Github repository: