学术报告:Symposium on Clustering and Prediction Intelligent Algorithm


报告一A Development of Interpretable Rule-Based Architecture Under Privacy Constraints: A Framework of Granular Computing

报告人 Prof. Witold PedryczInformation Science主编

单位阿尔伯塔大学(加拿大)

摘要In data analytics, system modeling, and decision-making models, the aspects of interpretability and explainability are of paramount relevance as emphasized in studies on explainable Artificial Intelligence (XAI). Those requirements are especially timely when the design of models has to be realized when considering strict requirements of privacy and security. We advocate that to efficiently address these challenges, it becomes beneficial to engage the fundamental framework of Granular Computing. It is demonstrated that a conceptualization of information granules can be conveniently carried out with the use of information granules (for example, fuzzy sets, sets, rough sets, and alike). We present a comprehensive discussion of information granules-oriented design of rule-based architectures and their interpretation by engaging an innovative mechanism of federated unsupervised learning using which information granules are constructed.

 

报告二:Objective-Domain Dual Decomposition: An Effective Approach to Optimizing Partially Differentiable Objective Functions

报告人张晓明教授IEEE Fellow           单位:香港浸会大学

摘要This paper addresses a class of optimization problems in which either part of the objective function is differentiable while the rest is nondifferentiable or the objective function is differentiable in only part of the domain. Accordingly, we propose a dual-decomposition-based approach that includes both objective decomposition and domain decomposition. In the former, the original objective function is decomposed into several relatively simple subobjectives to isolate the nondifferentiable part of the objective function, and the problem is consequently formulated as a multiobjective optimization problem (MOP). In the latter decomposition, we decompose the domain into two subdomains, that is, the differentiable and nondifferentiable domains, to isolate the nondifferentiable domain of the nondifferentiable subobjective. Subsequently, the problem can be optimized with different schemes in the different subdomains. We propose a population-based optimization algorithm, called the simulated water-stream algorithm (SWA), for solving this MOP. The SWA is inspired by the natural phenomenon of water streams moving toward a basin, which is analogous to the process of searching for the minimal solutions of an optimization problem. The proposed SWA combines the deterministic search and heuristic search in a single framework. Experiments show that the SWA yields promising results compared with its existing counterparts.

 

报告三:Evolutionary Mult-objective and Many-objective Optimization Algorithm Using Region Decomposition

报告人海林教授           单位:广东工业大学

摘要:该报告主要介绍在目标空间区域分解的思想,以及基于区域分解的进化多目标算法框架。给出几种在该框架下设计的一些进化多目标和超多目标优化算法,提出了不均衡多目标最优化问题,并从理论上分析了基于区域分解的进化多目标算法在处理这类问题时的优势。

 

报告四Design Automation of Intelligent Robotic Systems Based on Evolutionary Computation

报告人:范衠教授             单位:汕头大学

摘要The main reason why the performance of domestic robots is generally difficult to reach the same level of that of foreign countries mainly lies in the lack of systematic continuous optimization and design automation. How to form a framework of design automation of intelligent robotic systems is the main topic of this report. This report will mainly focus on multi-angle modeling methods of intelligent robotic systems, solving intelligent robot optimization problems by combining constrained multi-objective evolutionary algorithms and machine learning methods, and applying design automation methods to develop intelligent robots.

 

报告五:复杂网络建模、认知与优化

报告人:刘静教授(万人计划)             单位西安电子科技大学

摘要:复杂网络已成为研究复杂系统的有力工具。网络化系统往往是大型或巨型的复杂系统,需要处理海量信息。而要控制进而利用网络化系统首先要对其进行建模——构建出网络以刻画组成部分间的作用关系,进而对其特性进行认知——学习网络本质特性、结构特点、演化趋势、各种调整可能带来的结果等,进而才能对其进行优化——根据需要对复杂系统进行设计、控制、调整。本报告对复杂网络的建模、认知、优化进行全链条的介绍,并介绍基于进化算法的复杂网络建模、认知、优化方法。

 

报告六:基于集成学习预测的动态多目标优化方法及应用

报告人:王峰教授            单位武汉大学

摘要:动态多目标优化是目前多目标优化领域的一个研究热点,目前研究最为广泛的一种思路是通过预测策略指导算法搜索找到最优解。为了更好的应对不同类型的环境变化,克服单一预测模型在不同类型环境变化下的不稳定性,提出了一种基于集成学习的预测策略(Ensemble Learning-based Prediction Strategy, ELPS)帮助算法重初始化种群以适应新的环境。在ELPS中选取四种不同类型的预测模型作为基模型,包括基于种群的自回归模型、线性预测模型、基于拐点的自回归模型及随机初始化模型。当环境发生改变时,ELPS可以通过对历史种群进行训练得到一个强预测模型,并重新生成新种群,实现对环境变化的动态响应。通过ELPS的训练,可以有效提升预测的准确性并提高种群的多样性。为了进一步验证算法在实际问题的性能,选取带时间窗口的多目标动态车辆路径规划问题进行研究,进一步提出了一个基于ELPS的用以求解带时间窗口的多目标动态车辆路径规划问题的算法,选取了随机初始化策略模型、迁入模型和基于种群的预测模型作为基模型,可以实现对环境不同程度的波动的快速响应,加速种群的收敛速度。

 

报告七:Role-based Cooperation in Swarm Intelligence

报告人:夏学文教授          单位闽南师范大学

摘要It is a common phenomenon in human society that allocating different tasks to different people according to their capabilities (or roles). Many studies also verify that assigning different search strategies for different individuals can yield very favorable performance. Furthermore, multi-swarm techniques have been successfully applied in swarm intelligence algorithms (SI) since they can yield very pleasurable performance in keeping population diversity. Inspired by these observations, we integrate multiple roles into SI. In the proposed strategy, the entire population is split into multiple sub-swarms. During the evolutionary process, individuals in each sub-swarm adaptively select their own breeding strategies based on their own roles. Furthermore, the population is regrouped during the evolution process. Thus, different individuals in the same sub-swarm play different roles in a generation. Moreover, an individual may play different roles in different generations, Even a same individual may play different roles in different sub-swarms. Although the role-based cooperation offers a competitive performance testified by the extensive experiments, there are a few problems need to be further studied in future work. The one is the efficiency and effectiveness of the cooperation between different roles need to be in-depth analyzed. The other one is when and how to adjust the population size of different roles to satisfy various fitness landscapes and different evolution stages.

 

报告八:深度强化学习求解组合优化问题

报告人王甲海教授           单位:中山大学

摘要智能优化算法(启发式或者元启发式)的启发式规则需人工设计。如启发式规则能够通过机器学习方法,特别端到端的深度学习方法,来自动学习得到,就减少了人为设计启发式规则(算法)的工作量。这个习得的启发式算法,表示为一个训练好了的模型,可实现一次(离线)训练多次(在线)求解,针对没见过的问题实例,模型快速生成一个优质解,不需像传统智能优化算法那样,面对每一个问题的不同算例,都从初始解开始不断迭代寻优。讲座介绍深度强化学习求解组合优化问题的基本技术和方法。