Uniform-Circuit and Logarithmic-Space Approximations of Refined Combinatorial Optimization Problems

TL;DR

This paper explores the computational complexity and approximation algorithms for low-space and uniform-circuit classes of combinatorial optimization problems, introducing new classes and analyzing their relationships.

Contribution

It expands the framework for analyzing approximation algorithms within low-complexity classes, introduces new classes, and identifies complete problems and class collapses.

Findings

Identified new NLO problems in low-complexity classes.

Established approximation-preserving reductions among classes.

Demonstrated class collapses and separations under certain assumptions.

Abstract

A significant progress has been made in the past three decades over the study of combinatorial NP optimization problems and their associated optimization and approximate classes, such as NPO, PO, APX (or APXP), and PTAS. Unfortunately, a collection of problems that are simply placed inside the P-solvable optimization class PO never have been studiously analyzed regarding their exact computational complexity. To improve this situation, the existing framework based on polynomial-time computability needs to be expanded and further refined for an insightful analysis of various approximation algorithms targeting optimization problems within PO. In particular, we deal with those problems characterized in terms of logarithmic-space computations and uniform-circuit computations. We are focused on nondeterministic logarithmic-space (NL) optimization problems or NPO problems. Our study covers a wide range of optimization and approximation classes, dubbed as, NLO, LO, APXL, and LSAS as well as new classes NC1O, APXNC1, NC1AS, and AC0O, which are founded on uniform families of Boolean circuits. Although many NL decision problems can be naturally converted into NL optimization (NLO) problems, few NLO problems have been studied vigorously. We thus provide a number of new NLO problems falling into those low-complexity classes. With the help of NC1 or AC0 approximation-preserving reductions, we also identify the most difficult problems (known as complete problems) inside those classes. Finally, we demonstrate a number of collapses and separations among those refined optimization and approximation classes with or without unproven complexity-theoretical assumptions.