Escaping Local Optima: Constraint Weights vs. Value Penalties

Basharu, Muhammed; Arana, Inés; Ahriz, Hatem

doi:10.1007/978-1-84800-094-0_5

Cited by 4 publications

(2 citation statements)

References 4 publications

(4 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Incremental penalties are small and remain imposed until they are reset while temporary penalties are discarded immediately after they are used. The penalties on values approach has been shown to outperform the weights on constraints approach of escaping local optima (Basharu et al, 2007a).…”

Section: Related Workmentioning

confidence: 99%

Dynamic Agent Prioritisation with Penalties in Distributed Local Search

Sambo-Magaji¹,

Arana²,

Ahriz³

2013

Proceedings of the 5th International Conference on Agents and Artificial Intelligence

View full text Add to dashboard Cite

Distributed Constraint Satisfaction Problems (DisCSPs) solving techniques solve problems which are distributed over a number of agents.The distribution of the problem is required due to privacy, security or cost issues and, therefore centralised problem solving is inappropriate. Distributed local search is a framework that solves large combinatorial and optimization problems. For large problems it is often faster than distributed systematic search methods. However, local search techniques are unable to detect unsolvability and have the propensity of getting stuck at local optima. Several strategies such as weights on constraints, penalties on values and probability have been used to escape local optima. In this paper, we present an approach for escaping local optima called Dynamic Agent Prioritisation and Penalties (DynAPP) which combines penalties on variable values and dynamic variable prioritisation for the resolution of distributed constraint satisfaction problems. Empirical evaluation with instances of random, meeting scheduling and graph colouring problems have shown that this approach solved more problems in less time at the phase transition when compared with some state of the art algorithms. Further evaluation of the DynAPP approach on iteration-bounded optimisation problems showed that DynAPP is competitive.

show abstract

Section: Related Workmentioning

confidence: 99%

Dynamic Agent Prioritisation with Penalties in Distributed Local Search

Sambo-Magaji¹,

Arana²,

Ahriz³

2013

Proceedings of the 5th International Conference on Agents and Artificial Intelligence

View full text Add to dashboard Cite

show abstract

“…In contrast to fixing, some of the approaches learn to avoid local optima in the search process. [14] compare two strategies of escaping local optima: (a) assigning penalties to violated constraints (b) assign penalties to individual variable values participating in a constraint violation. Their results quantify the impact of penalties on the solution landscape.…”

Section: Introductionmentioning

confidence: 99%

Constraint Programming to Discover One-Flip Local Optima of Quadratic Unconstrained Binary Optimization Problems

Verma¹,

Lewis²

2021

Preprint

View full text Add to dashboard Cite

The broad applicability of Quadratic Unconstrained Binary Optimization (QUBO) constitutes a general-purpose modeling framework for combinatorial optimization problems and are a required format for gate array and quantum annealing computers. QUBO annealers as well as other solution approaches benefit from starting with a diverse set of solutions with local optimality an additional benefit. This paper presents a new method for generating a set of one-flip local optima leveraging constraint programming. Further, as demonstrated in experimental testing, analysis of the solution set allows the generation of soft constraints to help guide the optimization process.

show abstract

Getting What You Pay For: Is Exploration in Distributed Hill Climbing Really Worth it?

Smith

Mailler

2010

2010 IEEE/WIC/ACM International Conference on Web Intelligence and Intelligent Agent Technology

View full text Add to dashboard Cite

Abstract-The Distributed Stochastic Algorithm (DSA), Distributed Breakout Algorithm (DBA), and variations such as Distributed Simulated Annealing (DSAN), MGM-1, and DisPeL, are distributed hill-climbing techniques for solving large Distributed Constraint Optimization Problems (DCOPs) such as distributed scheduling, resource allocation, and distributed route planning. Like their centralized counterparts, these algorithms employ escape techniques to avoid getting trapped in local minima during the search process. For example, the best known version of DSA, DSA-B, makes hill-climbing and lateral escape moves, moves that do not impact the solution quality, with a single probability p. DSAN uses a similar scheme, but also occasionally makes a move that leads to a worse solution in an effort to find a better overall solution. Although these escape moves tend to lead to a better solutions in the end, the cost of employing the various strategies is often not well understood. In this work, we investigate the costs and benefits of the various escape strategies by empirically evaluating each of these protocols in distributed graph coloring and sensor tracking domains. Through our testing, we discovered that by reducing or eliminating escape moves, the cost of using these algorithms decreases dramatically without significantly affecting solution quality.

show abstract

Escaping Local Optima: Constraint Weights vs. Value Penalties

Cited by 4 publications

References 4 publications

Dynamic Agent Prioritisation with Penalties in Distributed Local Search

Dynamic Agent Prioritisation with Penalties in Distributed Local Search

Constraint Programming to Discover One-Flip Local Optima of Quadratic Unconstrained Binary Optimization Problems

Getting What You Pay For: Is Exploration in Distributed Hill Climbing Really Worth it?

Contact Info

Product

Resources

About