Intelligent human action recognition using an ensemble model of evolving deep networks with swarm-based optimization
作者：
Highlights：
• We propose an evolving ensemble deep network for action classification.
• A swarm intelligence algorithm is proposed to optimize network hyper-parameters.
• It incorporates search coefficients extracted from a 3D super-ellipse surface.
• Diverse elite signals are generated by sine, cosine and tanh crossover operators.
• Our systems depict a superior capability in action classification.
摘要
•We propose an evolving ensemble deep network for action classification.•A swarm intelligence algorithm is proposed to optimize network hyper-parameters.•It incorporates search coefficients extracted from a 3D super-ellipse surface.•Diverse elite signals are generated by sine, cosine and tanh crossover operators.•Our systems depict a superior capability in action classification.
论文关键词：Swarm intelligence,Evolutionary algorithm,Deep hybrid neural network,Ensemble classifier,Human action recognition
论文评审过程：Received 30 August 2020, Revised 7 December 2020, Accepted 2 March 2021, Available online 5 March 2021, Version of Record 9 March 2021.
论文官网地址：https://doi.org/10.1016/j.knosys.2021.106918
