With the fast development of artificial intelligence, a lot of translation methods and search methods have been proposed to address molecular optimization problems in drug design, which enables this field to achieve remarkable progress. However, existing methods still encounter great difficulties in addressing problems involving more than three properties, since these problems pose stiff challenges to translation methods and search methods in terms of acquiring high-quality training data and balancing multiple properties, respectively. In this paper, we propose an adaptive evolutionary optimization framework to address the many-property molecular optimization problems (namely MaOMO). MaOMO adaptively identifies the property with the largest improvement potential in each iteration, which generates high-quality molecules as efficiently as possible by devoting more efforts to the property. Besides, MaOMO adopts a dynamic selection strategy to select molecules with large property improvement, good property diversity, and structure diversity. We investigate the performance of MaOMO framework on both benchmark and practical molecular optimization tasks, which involve the simultaneous optimization of four or more properties. Experimental results show that the proposed framework is superior to five state-of-the-art competitors, which achieves a success rate improvement of more than 20% on practical optimization tasks.

Sparse Large-scale Multi-objective Optimization Problems (sparse LMOPs) widely exist in various optimization applications, such as neural network training, portfolio optimization, and feature selection of classification. Although numerous methods exist, automatically selecting efficient solving strategies for sparse LMOPs remains highly challenging. Given this, we propose a reinforcement learning assisted autonomous sparse multi-objective evolutionary algorithm, which aims to effectively utilize sparse knowledge for designing diversified genetic operators, and automatically select appropriate genetic operators for various problems or different situations within the same optimization process. Specifically, three sparsity-aware genetic operators are designed by utilizing sparsity statistic, sparsity clustering, and sparsity logic operation. They possess distinct advantages in terms of convergence speed, solution quality, and diversity. Furthermore, the utilization of deep Q-network enables the automatic selection of suitable operators for offspring reproduction based on the current sparse state of the population. The proposed algorithm is compared with five state-of-the-art algorithms on eight benchmark and three real-world problems. Experimental results demonstrate the superiority of the proposed algorithm and the effectiveness of the proposed sparse genetic operators for solving sparse LMOPs.