This paper focuses on the exponential stability of time-varying hybrid stochastic large-scale networks, in which a novel discrete-time observation control strategy with volatile control gain is imposed. The control gain exhibits sign-indefinite characteristics that follow certain volatile patterns. Then, the strategy can preferably respond to extrinsic perturbances and defend actuators and facilities by comparison with the position in constant gain. To quantify the volatile control gain, a concept of average volatile discrete-time observation control gain is proposed. In light of stochastic analysis theory and Lyapunov method, we establish improved stability criteria. In contrast to earlier results, the piecewise continuous, time-varying, and indefinite functions replace the constant coefficient of the upper bound estimation for the operator of the Lyapunov function, i.e., it can be positive or negative along time evolution, which shows wider applications of our results. Furthermore, considering the phenomenon of controller failure and packet losses, obtained results are applicable for analyzing the stabilization via intermittent discrete-time observation control and nonuniform discrete-time observation control, respectively. Finally, two applications to oscillator systems and single-link robot arms are presented and numerical simulations are exploited to illustrate the feasibility. Note to Practitioners—This paper is motivated by existing results on discrete-time observation control for the dynamics of hybrid stochastic large-scale networks. The existing results mainly require continuous updates to the control signals while maintaining a constant control gain. This is challenging to achieve on digital computers and often does not effectively respond to external disturbances. This paper designs a new hybrid control called volatile discrete-time observation control, and a concept of the control gain is proposed to quantify the volatile control gain. Specifically, the obtained results also apply to scenarios involving control failures and packet loss. The results are applied to oscillator systems and single-link robot arms, demonstrating their effectiveness and it is expected that the proposed approach can be extended to more practical physical engineering systems.

Aspect-based sentiment classification aims to identify sentiment polarity expressed towards a given opinion target in a sentence. The sentiment polarity of the target is not only highly determined by sentiment semantic context but also correlated with the concerned opinion target. Existing works cannot effectively capture and store the inter-dependence between the opinion target and its context. To solve this issue, we propose a novel model of Attentive Neural Turing Machines (ANTM). Via interactive read-write operations between an external memory storage and a recurrent controller, ANTM can learn the dependable correlation of the opinion target to context and concentrate on crucial sentiment information. Specifically, ANTM separates the information of storage and computation, which extends the capabilities of the controller to learn and store sequential features. The read and write operations enable ANTM to adaptively keep track of the interactive attention history between memory content and controller state. Moreover, we append target entity embeddings into both input and output of the controller in order to augment the integration of target information. We evaluate our model on SemEval2014 dataset which contains reviews of Laptop and Restaurant domains and Twitter review dataset. Experimental results verify that our model achieves state-of-the-art performance on aspect-based sentiment classification.