Given the original paper (https://arxiv.org/pdf/1809.02864.pdf), I would like to implement the Accelegrad algorithm for which I report the pseudocode of the paper:

enter image description here

In the pseudocode, the authors refer to a compact convex set $K$ of diameter $D$. The question is whether I can know such elements. I think that they are theoretical conditions to satisfy some theorems. The problem is that the diameter $D$ is used in the learning rate and also the convex set $K$ is used to perform the projection of the gradient descent. How can I proceed?



You must log in to answer this question.