学术报告第201762期

摘要：There are many swarm   intelligence algorithms in literature. Each has its own strength and weakness   because of its own framework. In this talk, a unified swarm intelligence   framework will be presented. First, the necessity of having developmental   learning embedded in a swarm intelligence algorithm is confirmed by briefly   considering brain evolution, brain development, brainstorming process, etc.   Secondly, a framework of a developmental swarm intelligence algorithm, which   contains capacity developing stage and capability learning stage, is given to   help understand developmental swarm intelligence (DSI) algorithms, and to   guide to design and/or implement any new developmental swarm intelligence   algorithm and/or any developmental evolutionary algorithm. Following DSI,   innovation is discussed and an innovation-inspired optimization (IO)   algorithm is designed and developed. Finally, by combing the DSI and IO   algorithm together, a unified swarm intelligence framework is proposed, which   contains capacity developing stage and capability learning stage and with   three search operators in its capability learning stage to mimicking the   three levels of innovations，i.e., at the learning   level, at the emulating level, and at the exploring level, respectively.