面向稀疏多目标优化问题的进化算法研究

Sparse multi-objective optimization problems widely exist in many fields including machine learning, data mining, and network science. Although evolutionary algorithms have shown promising performance in solving a variety of complex multi-objective optimization problems, little work has been dedicated to solving sparse multi-objective optimization problems. To address this issue, this project studies the evolutionary algorithms for sparse multi-objective optimization problems, the main research contents of which include: (1) developing a general evolutionary algorithm for sparse multi-objective optimization problems, to solve the sparse multi-objective optimization problems in various fields; (2) designing benchmark problems and a performance metric for sparse multi-objective optimization, to provide a systematic assessment method for sparse multi-objective evolutionary algorithms; (3) applying the proposed algorithm to several sparse multi-objective optimization problems in real-world applications, and establishing a simulation platform for sparse multi-objective optimization, to provide a complete optimization tool for solving this type of problems. This project can not only provide an effective sparse multi-objective evolutionary algorithm for evolutionary computation, but also offer a practical optimizer for the sparse multi-objective optimization problems in various fields.

稀疏多目标优化问题广泛存在于机器学习、数据挖掘、网络科学等领域中。虽然进化算法在求解许多复杂多目标优化问题上取得了良好的效果，但关于求解稀疏多目标优化问题的研究仍然很少。为此，本项目研究面向稀疏多目标优化问题的进化算法，主要研究内容包括：（1）开发针对稀疏多目标优化问题的通用进化算法，用于求解不同领域中的稀疏多目标优化问题；（2）设计用于稀疏多目标优化的测试问题及性能指标，为面向稀疏多目标优化问题的进化算法提供系统的测试方法；（3）将拟开发的算法应用于多个实际中的稀疏多目标优化问题，并建立稀疏多目标优化仿真平台，为该类问题的求解提供一套完整的优化工具。本项目的研究不仅为进化计算领域引入求解稀疏多目标优化问题的高效算法，同时也为不同领域中的稀疏多目标优化问题提供一种实用的优化方法。

本项目研究基于进化算法的稀疏多目标优化问题求解，通过开发面向稀疏多目标优化问题的通用进化算法，为进化计算领域引入一类新型进化算法，并为许多研究领域中的稀疏多目标优化问题提供一种高效的优化方法。本项目的具体研究目标包括面向稀疏多目标优化问题的通用进化算法、面向稀疏多目标优化问题的问题集和性能指标、稀疏多目标进化算法的应用及平台建设。本项目的研究目标均已完成，累积发表学术论文24篇（其中IEEE/ACM汇刊论文20篇），申请发明专利8项，获得科研奖励7项，作大会特邀报告3次，培养研究生2人。本项目所开发的理论及方法弥补了现有稀疏多目标优化领域的不足，并成功应用于放疗计划、疾病模块检测、深度学习、路径规划、社团检测、故障诊断等领域，具有一定的实际价值。