Stability Analysis of Recurrent Neural Networks With Time-Varying Delay Based on a Flexible Negative-Determination Quadratic Function Method
Guoqiang Tan, Zhanshan Wang
Abstract
This brief investigates the stability problem of recurrent neural networks (RNNs) with time-varying delay. First, by introducing some flexibility factors, a flexible negative-determination quadratic function method is proposed, which contains some existing methods and has less conservatism. Second, some integral inequalities and the flexible negative-determination quadratic function method are used to give an accurate upper bound of the Lyapunov-Krasovskii functional (LKF) derivative. As a result, a less conservative stability criterion of delayed RNNs is derived, whose effectiveness and superiority are finally illustrated through two numerical examples.
Topics & Concepts
Quadratic equationArtificial neural networkControl theory (sociology)Stability (learning theory)Quadratic functionFunction (biology)Activation functionComputer scienceRecurrent neural networkMathematicsAlgorithmArtificial intelligenceMachine learningControl (management)BiologyEvolutionary biologyGeometryNeural Networks and ApplicationsNeural Networks Stability and SynchronizationMachine Learning and ELM