Generation of tests for programming challenge tasks using multi-objective optimization

Buzdalov, Maxim; Buzdalova, Arina; Petrova, Irina

doi:10.1145/2464576.2482746

Cited by 12 publications

(5 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…We consider optimization problems where all objectives, including the target objective, are calculated during one processing of an individual [3,4]. Therefore, evaluation of reward does not increase the number of objective evaluations.…”

Section: Related Work: Ea+rl Methodsmentioning

confidence: 99%

See 1 more Smart Citation

Selection of Auxiliary Objectives with Multi-Objective Reinforcement Learning

Buzdalova

Matveeva

Korneev

2015

Proceedings of the Companion Publication of the 2015 Annual Conference on Genetic and Evolutionary Computation

Self Cite

View full text Add to dashboard Cite

Efficiency of evolutionary algorithms may be increased using multi-objectivization. Multi-objectivization is performed by adding some auxiliary objectives. We consider selection of these objectives during a run of an evolutionary algorithm.One of the selection methods is based on reinforcement learning. There are several types of rewards previously used in reinforcement learning for adjusting of evolutionary algorithms. However, there is no superior reward. At the same time, reinforcement learning itself may be enhanced by multi-objectivization. So we propose a method for selection of auxiliary objectives based on multi-objective reinforcement learning, where the reward is composed of the previously used single rewards. Hence, we have double multiobjectivization: several rewards are involved in selection of several auxiliary objectives.We run the proposed method on different benchmark problems and compare it with a conventional evolutionary algorithm and a method based on single-objective reinforcement learning. Multi-objective reinforcement shows competitive behavior and is especially useful in the case when we do not know in advance which of the single rewards is efficient.

show abstract

Section: Related Work: Ea+rl Methodsmentioning

confidence: 99%

“…The Avg reward is the difference between the averaged target fitness of the current generation and the previous generation [3]:…”

Section: Rewardsmentioning

confidence: 99%

Selection of Auxiliary Objectives with Multi-Objective Reinforcement Learning

Buzdalova

Matveeva

Korneev

2015

Proceedings of the Companion Publication of the 2015 Annual Conference on Genetic and Evolutionary Computation

Self Cite

View full text Add to dashboard Cite

show abstract

“…Число вычислений ФП взято из работы [7]. Для задач ran20, ran50, euc50 число вычислений ФП равно 5·10 5 , а для задач euc100, 464 kroB100 -2·10 6 . Алгоритм ε-жадного Q-обучения применялся со следующими значениями параметров: α = 0,6, γ = 0,01, ε = 0,3.…”

Section: экспериментальное исследование метода Ea+rlunclassified

“…В предложенном методе для выбора ФП впервые было применено обучение с подкреплением [2,3]. Эффективность метода была подтверждена теоретически [4] и экспериментально путем применения его для решения ряда задач оптимизации [5,6].…”

Section: Introductionunclassified

Adaptive selection of auxiliary objectives in multiobjective evolutionary algorithms

Petrova¹,

Buzdalova²,

Shalyto³

2016

Naučno-teh. vestn. inf. tehnol. meh. opt.

View full text Add to dashboard Cite

“…The method was shown to be efficient, both empirically (for a number of problems, including a real-world application [1]) and theoretically (for a model OneMax problem [2]). Thus in this paper we propose a method of evolutionary operators selection based on our previous ideas, which seems to be promising.…”

Section: Introductionmentioning

confidence: 99%

Selecting evolutionary operators using reinforcement learning

Buzdalova

Kononov

Buzdalov

2014

Proceedings of the Companion Publication of the 2014 Annual Conference on Genetic and Evolutionary Computation

Self Cite

View full text Add to dashboard Cite

In evolutionary optimization, it is important to use efficient evolutionary operators, such as mutation and crossover. But it is often difficult to decide, which operator should be used when solving a specific optimization problem. So an automatic approach is needed. We propose an adaptive method of selecting evolutionary operators, which takes a set of possible operators as input and learns what operators are efficient for the considered problem. One evolutionary algorithm run should be enough for both learning and obtaining suitable performance. The proposed EA+RL(O) method is based on reinforcement learning. We test it by solving H-IFF and Travelling Salesman optimization problems. The obtained results show that the proposed method significantly outperforms random selection, since it manages to select efficient evolutionary operators and ignore inefficient ones.

show abstract

Generation of tests for programming challenge tasks using multi-objective optimization

Cited by 12 publications

References 21 publications

Selection of Auxiliary Objectives with Multi-Objective Reinforcement Learning

Selection of Auxiliary Objectives with Multi-Objective Reinforcement Learning

Adaptive selection of auxiliary objectives in multiobjective evolutionary algorithms

Selecting evolutionary operators using reinforcement learning

Contact Info

Product

Resources

About