Ubiquitous algorithms in convex optimization generate self-contracted sequences

2020/03/09 by Böhm, Axel, Daniilidis, Aris · 1 citation
#FOS: Mathematics #Optimization and Control (math.OC)

paper · doi:10.48550/arxiv.2003.04201

Abstract

In this work we show that various algorithms, ubiquitous in convex optimization (e.g. proximal-gradient, alternating projections and averaged projections) generate self-contracted sequences \x_k\_k∈ℕ. As a consequence, a novel universal bound for the length (∑_{k≥ 0}\Vert x_k+1-x_k\Vert) can be deduced. In addition, this bound is independent of both the concrete data of the problem (sets, functions) as well as the stepsize involved, and only depends on the dimension of the space.

Cited by

• Characterization of regularity via variational stability of alternating projections sequences

Related

• Revisiting the Polyak step size
• A General Analysis Framework of Lower Complexity Bounds for Finite-Sum Optimization
• Enumerative Lattice Algorithms in Any Norm via M-Ellipsoid Coverings
• A Lower Bound for the Optimization of Finite Sums
• On approximately Convex and Affine Sequences