no code implementations • 24 Jun 2021 • Patrick R. Johnstone, Jonathan Eckstein, Thomas Flynn, Shinjae Yoo
We present a new, stochastic variant of the projective splitting (PS) family of algorithms for monotone inclusion problems.
2 code implementations • 24 Feb 2019 • Patrick R. Johnstone, Jonathan Eckstein
In the convex optimization context, cocoercivity is equivalent to Lipschitz differentiability.
no code implementations • 17 Sep 2018 • Patrick R. Johnstone, Jonathan Eckstein
A recent innovation in projective splitting algorithms for monotone operator inclusions has been the development of a procedure using two forward steps instead of the customary proximal steps for operators that are Lipschitz continuous.
no code implementations • 11 Jun 2018 • Patrick R. Johnstone, Jonathan Eckstein
Second, for strongly monotone inclusions, strong convergence is established as well as an ergodic $O(1/\sqrt{k})$ convergence rate for the distance of the iterates to the solution.
1 code implementation • 19 Mar 2018 • Patrick R. Johnstone, Jonathan Eckstein
Forward steps can be used for any Lipschitz-continuous operators provided the stepsize is bounded by the inverse of the Lipschitz constant.
no code implementations • ICML 2017 • Jonathan Eckstein, Noam Goldberg, Ai Kagawa
We describe a learning procedure enhancing L1-penalized regression by adding dynamically generated rules describing multidimensional “box” sets.