Shannon-and von neumann-entropy regularizations of linear and semidefinite programs

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Chhatoi, Saroj Prasad, Lasserre, Jean B
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910905040633856
author Chhatoi, Saroj Prasad
Lasserre, Jean B
author_facet Chhatoi, Saroj Prasad
Lasserre, Jean B
contents We consider the LP in standard form min {c T x\,: Ax = b; x $\ge$ 0} and inspired by $ε$-regularization in Optimal Transport, we introduce its $ε$-regularization ''min {c T x + $ε$ f (x)\,: Ax = b; x $\ge$ 0}'' via the (convex) Boltzmann-Shannon entropy f (x)\,:= i x i ln x i . We also provide a similar regularization for the semidefinite program ''min {Tr(C $\bullet$ X)\,: A(X) = b; X 0}'' but with now the so-called Von Neumann entropy, as in Quantum Optimal Transport. Importantly, both are not barriers of the LP and SDP cones respectively. We show that this problem admits an equivalent unconstrained convex problem max $λ$$\in$R m G$ε$($λ$) for an explicit concave differentiable function G$ε$ in dual variables $λ$ $\in$ R m . As $ε$ goes to zero, its optimal value converges to the optimal value of the initial LP. While it resembles the log-barrier formulation of interior point algorithm for the initial LP, it has a distinguishing advantage. Namely for fixed $λ$, G$ε$($λ$) is obtained as a minimization over the whole space x $\in$ R d (and not over x $\ge$ 0) to still obtain a nonnegative solution x($λ$) $\ge$ 0, whence an explicit form of G$ε$ very useful for its unconstrained maximization over R m .
format Preprint
id arxiv_https___arxiv_org_abs_2503_23815
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Shannon-and von neumann-entropy regularizations of linear and semidefinite programs
Chhatoi, Saroj Prasad
Lasserre, Jean B
Optimization and Control
We consider the LP in standard form min {c T x\,: Ax = b; x $\ge$ 0} and inspired by $ε$-regularization in Optimal Transport, we introduce its $ε$-regularization ''min {c T x + $ε$ f (x)\,: Ax = b; x $\ge$ 0}'' via the (convex) Boltzmann-Shannon entropy f (x)\,:= i x i ln x i . We also provide a similar regularization for the semidefinite program ''min {Tr(C $\bullet$ X)\,: A(X) = b; X 0}'' but with now the so-called Von Neumann entropy, as in Quantum Optimal Transport. Importantly, both are not barriers of the LP and SDP cones respectively. We show that this problem admits an equivalent unconstrained convex problem max $λ$$\in$R m G$ε$($λ$) for an explicit concave differentiable function G$ε$ in dual variables $λ$ $\in$ R m . As $ε$ goes to zero, its optimal value converges to the optimal value of the initial LP. While it resembles the log-barrier formulation of interior point algorithm for the initial LP, it has a distinguishing advantage. Namely for fixed $λ$, G$ε$($λ$) is obtained as a minimization over the whole space x $\in$ R d (and not over x $\ge$ 0) to still obtain a nonnegative solution x($λ$) $\ge$ 0, whence an explicit form of G$ε$ very useful for its unconstrained maximization over R m .
title Shannon-and von neumann-entropy regularizations of linear and semidefinite programs
topic Optimization and Control
url https://arxiv.org/abs/2503.23815