Chips-n-Salsa: A Java Library of Customizable, Hybridizable, Iterative, Parallel, Stochastic, and Self-Adaptive Local Search Algorithms

Abstract

Discrete optimization problems are often of practical real-world importance as well as computationally intractable. For example, the traveling salesperson, bin packing, and longest common subsequence problems are NP-Hard, as is resource constrained scheduling, and many single-machine scheduling problems (Garey & Johnson, 1979). Polynomial time algorithms for such problems are unlikely to exist, and the best known algorithms that guarantee optimal solutions have a worst-case exponential runtime. It is thus common to use stochastic local search and other metaheuristics (Gonzalez, 2018). Stochastic local search algorithms begin at a random search state, and apply a sequence of neighbor transitions to nearby search states. This includes perturbative (Hoos & Stützle, 2018) algorithms like simulated annealing (Delahaye, Chaimatanan, & Mongeau, 2019) and hill climbers (Hoos & Stützle, 2018), where each search state is a complete candidate feasible solution, and a mutation operator makes a small random modification to move to another local candidate solution; and also includes constructive (Hoos & Stützle, 2018) algorithms like stochastic samplers (

Cycle Mutation: Evolving Permutations via Cycle Induction

“…Our Java implementations of cycle mutation are integrated into an open source library, Chips-n-Salsa [30], and our Java implementations of cycle distance, k-cycle distance, and cycle edit distance are integrated into the open source JavaPermutationTools [31] library. By disseminating the implementations in open source libraries, we hope to contribute not only to the research literature, but also to the state of practice.…”

“…A secondary aim of this research is to enable reproducibility [29], as well as to advance the state of practice. Therefore, we integrate our Java implementation of cycle mutation into the open source Chips-n-Salsa library [30], and cycle distance, k-cycle distance, and cycle edit distance into the open source JavaPermutationTools library [31]. Chips-n-Salsa is a library for stochastic and adaptive local search as well as EAs.…”

ALNS: a Python implementation of the adaptive large neighbourhood search metaheuristic

“…As such, these libraries provide limited functionality for single-trajectory algorithms like ALNS. Among libraries that focus on single-trajectory algorithms are Paradiseo (Dreo et al, 2021), and Chips-n-Salsa (Cicirello, 2020). Neither provides an implementation of the ALNS algorithm out of the box.…”

Optimizing the Modified Lam Annealing Schedule

“…We implemented both the original Modified Lam and our optimized version within an open-source Java library, Chips-n-Salsa (release 2.2.0), of stochastic local search algorithms (Cicirello, 2020). The source code of Chips-n-Salsa is maintained on GitHub 1 , and prebuilt releases are regularly deployed to Maven Central 2 as well as GitHub Packages 3 where from either of which the library can easily be imported into projects using common build tools such as Maven or Gradle (among others).…”

“…In this paper, in Section 3, we propose an optimized variant of Modified Lam, an adaptive annealing schedule. We have integrated this optimized Modified Lam into the open-source Java library, Chips-n-Salsa (Cicirello, 2020), which is a library of stochastic local search algorithms. The source code of the Chipsn-Salsa library is maintained on GitHub (https: //github.com/cicirello/Chips-n-Salsa), and releases can be downloaded and installed from Maven Central (https://search.maven.org/artifact/ org.cicirello/chips-n-salsa).…”