1. Synchronization control for Markov jump neural networks subject to HMM observation and partially known detection probabilities.
- Author
-
Li, Feng, Song, Shuai, Zhao, Jianrong, Xu, Shengyuan, and Zhang, Zhengqiang
- Subjects
- *
ARTIFICIAL neural networks , *HIDDEN Markov models , *PROBABILITY theory , *SYNCHRONIZATION , *ATTENTION control , *MARKOV spectrum - Abstract
• An hidden Markov model is used to avoid some imperfect hypotheses when handling the synchronization control issue for MJNNs. • The HMM is with partial known detection probabilities, which can remedy the disadvantage of some previous works. • Activation function dividing method is used to get less conservative result for the synchronization control issue for MJNNs. This paper pays attention to the synchronization control issue for Markov jump neural networks with partial information on system modes (or called Markov states), which leads to the case that the system modes cannot be directly accessed. An hidden Markov model (HMM)-based detector with partially known detection probabilities is employed to detect the system modes. With the help of the HMM and an activation function dividing method, a less conservative controller design technique is established. The designed HMM-based controller can be converted to mode-independent/-dependent one by suitably adjusting some design parameters. Finally, the availability of the established HMM-based controller design technique is verified by an illustrative example. [ABSTRACT FROM AUTHOR]
- Published
- 2019
- Full Text
- View/download PDF