Modified Choice Function Heuristic Selection for the Multidimensional Knapsack Problem

Drake, John H.; Özcan, Ender; Burke, Edmund K.

doi:10.1007/978-3-319-12286-1_23

Cited by 14 publications

(7 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In [57] a hyper-heuristic with a parameter free choice function strategy has been applied on pairwise test generation. In [58] a modified choice function heuristic selection strategy has been proposed for the multidimensional knapsack problem. In [59], a Choice Function-based Constructive Hyper-Heuristic is used for generating personalized healthy menu recommendations.…”

Section: Additive Reinforcement Learning Hyper-heuristics Behavior Anmentioning

confidence: 99%

A multilevel synergy Thompson sampling hyper-heuristic for solving Max-SAT

Lassouaoui

Boughaci

Benhamou³

2019

IDT

View full text Add to dashboard Cite

Hyper-heuristics are high-level methods, used to solve various optimization problems. Some of them are capable of learning and adapting their behavior throughout the solving process. Selection hyper-heuristics evaluate low-level heuristics and determine which of them to be applied at a given point in the search process. However, it has been shown that the additive learning process becomes inefficient in hard problems where the probability of fitness improvement is less than 1 2. Other alternative learning mechanisms have been proposed however they don't take into account the synergy between the low-level heuristics. Moreover they haven't been tested on large NP-hard problems. In this work, we propose a new hyper-heuristic which we called the Multilevel Synergy Thompson Sampling Hyper-Heuristic. The proposed method includes both the probabilistic learning mechanism and the multilevel paradigm. The latter refers to the process of creating hierarchically smaller sub-problems from large problem instances. The proposed hyper-heuristic is applied on very large industrial Max-SAT instances from the latest Max-SAT competition. The numerical results are promising and demonstrate the benefits of our method. The proposed method outperforms the other four types of hyper-heuristics: Random, Choice Function, Stochastic Choice Function and the simple Thompson Sampling Hyper-Heuristics.

show abstract

Section: Additive Reinforcement Learning Hyper-heuristics Behavior Anmentioning

confidence: 99%

A multilevel synergy Thompson sampling hyper-heuristic for solving Max-SAT

Lassouaoui

Boughaci

Benhamou³

2019

IDT

View full text Add to dashboard Cite

show abstract

“…GAGP is compared with the standard GA algorithm and other GAGP is compared to the following constructive approaches : PECH (Primal Effective Capacity Heuristic) [26]; MAG [27]; VZ [28]; PIR (Pirkul 1987) and SCE (Shuffled Complex Evolution) [29]. GAGP is also compared to the following improvement approaches : CB [25]; NR (P) (New Reduction (Pirkul)) [30] and MCF (Modified Choice Function -Late Acceptance Strategy) [31]. The comparison is shown in…”

Section: G Comparison With the Literaturementioning

confidence: 99%

Knowledge-based Genetic Algorithm for the 0–1 Multidimensional Knapsack Problem

Rezoug

Bader-El-Den

Boughaci

2017

2017 IEEE Congress on Evolutionary Computation (CEC)

View full text Add to dashboard Cite

This paper presents an improved version of Genetic Algorithm (GA) to solve the 0-1 Multidimensional Knapsack Problem (MKP01), which is a well-known NP-hard combinatorial optimisation problem. In combinatorial optimisation problems, the best solutions have usually a common partial structure. For MKP01, this structure contains the items with a high values and low weights. The proposed algorithm called Genetic Algorithm Guided by Pretreatment information (GAGP) calculates these items and uses this information to guide the search process. Therefore, GAGP is divided into two steps, in the first, a greedy algorithm based on the efficiency of each item determines the subset of items that are likely to appear in the best solutions. In the second, this knowledge is utilised to guide the GA process. Strategies to generate the initial population and calculate the fitness function of the GA are proposed based on the pretreatment information. Also, an operator to update the efficiency of each item is suggested. The pretreatment information has been investigated using the CPLEX deterministic optimiser. In addition, GAGP has been examined on the most used MKP01 data-sets, and compared to several other approaches. The obtained results showed that the pretreatment succeeded to extract the most part of the important information. It has been shown, that GAGP is a simple but very competitive solution.

show abstract

“…Burke et al [6,7] offered a more general definition covering the two main classes of hyper-heuristics, selection hyper-heuristics (e.g. [8,9]) and generation hyper-heuristics (e.g. [10,11]):…”

Section: Introductionmentioning

confidence: 99%

“…Hyper-heuristics have been applied successfully to a variety of problems such as bin packing [12], dynamic environments [13], examination timetabling [14,15], multidimensional knapsack problem [9,16], nurse scheduling [14], production scheduling [17], sports scheduling [18] and vehicle routing [19].…”

Section: Introductionmentioning

confidence: 99%

Two Frameworks for Cross-Domain Heuristic and Parameter Selection Using Harmony Search

Dempster¹,

Drake²

2015

Advances in Intelligent Systems and Computing

Self Cite

View full text Add to dashboard Cite

Harmony Search is a metaheuristic technique for optimizing problems involving sets of continuous or discrete variables, inspired by musicians searching for harmony between instruments in a performance. Here we investigate two frameworks, using Harmony Search to select a mixture of continuous and discrete variables forming the components of a Memetic Algorithm for cross-domain heuristic search. The first is a single-point based framework which maintains a single solution, updating the harmony memory based on performance from a fixed starting position. The second is a population-based method which co-evolves a set of solutions to a problem alongside a set of harmony vectors. This work examines the behaviour of each framework over thirty problem instances taken from six different, real-world problem domains. The results suggest that population co-evolution performs better in a time-constrained scenario, however both approaches are ultimately constrained by the underlying metaphors.

show abstract

Modified Choice Function Heuristic Selection for the Multidimensional Knapsack Problem

Cited by 14 publications

References 24 publications

A multilevel synergy Thompson sampling hyper-heuristic for solving Max-SAT

A multilevel synergy Thompson sampling hyper-heuristic for solving Max-SAT

Knowledge-based Genetic Algorithm for the 0–1 Multidimensional Knapsack Problem

Two Frameworks for Cross-Domain Heuristic and Parameter Selection Using Harmony Search

Contact Info

Product

Resources

About