1 paper · 1 filter
Vincent Lemaire, Gilles Pagès, Christian Yeo
We introduce a new class of neural networks designed to be convex functions of their inputs, leveraging the principle that any convex function can be represented as the supremum of…