Coupled Oscillator Based Potts Machine to Solve Combinatorial Optimization Problems
Loading...
Author(s)
Babu, Nithin
Advisor(s)
Editor(s)
Collections
Supplementary to:
Permanent Link
Abstract
Combinatorial Optimization Problems (COP) represent a class of NP-Hard problems with diverse practical applications. These problems involve the partitioning of sets into discrete subsets while adhering to specific constraints. Among COPs, the max-3-cut problem stands out, presenting a scenario where a set is partitioned into three subsets with the objective of maximizing inter-set edge connections. Traditional approaches to address max-3-cut problems often resort to algorithms based on heuristic methods, simulated annealing, etc. to be run on conventional computers. However, these methods are plagued by extended computation times and substantial energy consumption. In response, physics-based solutions have been extensively explored, particularly Ising machines, which leverage Ising models. Although Ising machines offer notable energy and time savings, their hardware requirements introduce significant performance overheads when applied to max-3-cut problems. A more direct and energy-efficient alternative involves leveraging a hardware solver based on the Potts model. This research introduces a novel architecture for a coupled oscillator-based Potts machine (OPM) tailored to tackle max-3-cut problems. It also outlines the considerable advantages of employing an OPM over an oscillator Ising machine (OIM) for max-3-cut problem resolution. Finally, the study conducts a comparative analysis between the OPM and a state-of-the-art heuristic algorithm (MOH) for solving max-3-cut problems on large Gset networks, revealing a {10}^7x reduction in runtime while maintaining solution quality above 90%.
Sponsor
Date
2024-04-29
Extent
Resource Type
Text
Resource Subtype
Thesis