首页>
外国专利>
A distributed adaptive task-switching and learning based on clustering for secure multi-task estimation algorithm over adversarial multi-task network
A distributed adaptive task-switching and learning based on clustering for secure multi-task estimation algorithm over adversarial multi-task network
#$%^&*AU2019101411A420200102.pdf#####Abstract: A distributed adaptive task-switching and learning based on clustering algorithm for secure estimation over adversarial multi-task network. The algorithm consists of the following steps: false data injection attack detecting; resilient adaption for abnormal information; information sharing over multi-task network with local reliable neighbors according to the information credibility; adaptive pseudo clustering for selfish agents via an adaptive clustering threshold; information credibility updating; combining estimates with the trust neighbors; adaptive task-switching ; probability of sharing genuine information updating ,which all enhanced selfish agents' secure and effective information sharing that governed by our established selfish information sharing model. The complete algorithm set forth in the present invention performs more securer distributed estimation and more effective multi-task optimization, which mainly focus on selfish agents' dishonest behaviors while sharing information, especially for the selfish agents located in a deteriorating multi-task network, where the cost of sharing genuine information internally is generally high, while the external of the adversarial multi-task network is inevitably under malicious attacks. Experimental results indicate that the proposed algorithm is competitive to centralized estimation methods and existing distributed methods over adversarial multi-task network.1/4 Figi. N dk0U~ ~~*1~~~5 -~Ark Fig2. (a)(b
展开▼
