The performance of multiobjective algorithms varies across problems, making it hard to develop new algorithms or apply existing ones to new problems. To simplify the development and application of new multiobjective a...
详细信息
ISBN:
(纸本)9781450392372
The performance of multiobjective algorithms varies across problems, making it hard to develop new algorithms or apply existing ones to new problems. To simplify the development and application of new multiobjective algorithms, there has been an increasing interest in their automatic design from component parts. These automatically designed metaheuristics can outperform their human-developed counterparts. However, it is still uncertain what are the most influential components leading to their performance improvement. This study introduces a new methodology to investigate the effects of the final configuration of an automatically designed algorithm. We apply this methodology to a well-performing Multiobjective Evolutionary algorithm Based on Decomposition (MOEA/D) designed by the irace package on nine constrained problems. We then contrast the impact of the algorithm components in terms of their Search Trajectory Networks (STNs), the diversity of the population, and the hypervolume. Our results indicate that the most influential components were the restart and update strategies, with higher increments in performance and more distinct metric values. Also, their relative influence depends on the problem difficulty: not using the restart strategy was more influential in problems where MOEA/D performs better;while the update strategy was more influential in problems where MOEA/D performs the worst.
Per-Instance algorithm Selection and automatic algorithm configuration have recently gained important interests. However, these approaches face many limitations. For instance, the performance of these methods is deepl...
详细信息
ISBN:
(纸本)9781450367486
Per-Instance algorithm Selection and automatic algorithm configuration have recently gained important interests. However, these approaches face many limitations. For instance, the performance of these methods is deeply influenced by factors like the accuracy of the underlying prediction model, features space correlation, incomplete performance space for new instances, instances sampling and many others. In this paper, an effort to address such limitations is described. Indeed, we propose a cooperative architecture, labeled as the "SAPIAS" concept, composed of a self-adaptive online algorithm Selection system and an offline automatic algorithm configuration system, working together in order to deliver the most accurate performance. Additionally, SAPIAS is proposed as a methodic concept that the metaheuristics community might adopt to fill in the gap between theory and practice in the field, by providing for theoreticians the ability to continuously analyze the evolution of the problems characteristics and the behavior of the solving techniques as well as providing a ready to use solving framework for practitioners.
In this paper we address the non-permutation flow shop scheduling problem, a more general variant of the flow shop problem in which the machines can have different sequences of jobs. We aim to minimize the total compl...
详细信息
In this paper we address the non-permutation flow shop scheduling problem, a more general variant of the flow shop problem in which the machines can have different sequences of jobs. We aim to minimize the total completion time. We propose a template to generate iterated greedy algorithms, and use an automatic algorithm configuration to obtain efficient methods. This is the first automated approach in the literature for the non-permutation flow shop scheduling problem. The algorithms start by building a high-quality permutation solution, which is then improved during a second phase that generates non-permutation solutions by changing the job order on some machines. The obtained algorithms are evaluated against two well-known benchmarks from the literature. The results show that they can find better schedules than the state-of-the-art methods for both the permutation and non-permutation flow shop scheduling problems in comparable experimental conditions, as evidenced by comprehensive computational and statistical testing. We conclude that using non-permutation schedules is a viable alternative to reduce the total completion time that production managers should consider.
The parameter configuration problem consists of finding a parameter configuration that gives a particular algorithm the best performance. This paper introduces a new multi-phase tuner based on the iterated local searc...
详细信息
The parameter configuration problem consists of finding a parameter configuration that gives a particular algorithm the best performance. This paper introduces a new multi-phase tuner based on the iterated local search meta-heuristic. This tuner addresses the parameter configuration problem for deterministic MILP solvers that are used to solve challenging industrial optimization problems. Further, the proposed tuner offers a new search strategy based on three ideas. First, instead of tuning in the entire configuration space induced by the parameter set, the multi-phase tuner focuses on a small parameter pool that is dynamically enriched with new promising parameters. Second, it leverages the gathered knowledge during the search using statistical learning to forbid less promising parameter combinations. Third, it tunes on a single instance provided by earlier clustering of MILP instances. A computational study on the widely-used commercial solver CPLEX with instances from the MIPLIB library and a real large-scale optimization problem highlights the promising potential of the tuner.
暂无评论