Novel Tabu learning neuron model with variable activation gradient and its application to secure healthcare

Abstract

Currently, the latest advances in artificial neural networks have deeply affected various aspects of the general public. To this end, a new Tabu Learning Neuron (TLN) model with variable activation gradients is proposed in this paper. Specifically, its kinetic behaviors and intrinsic properties are investigated by means of a two-parameter Lyapunov exponential spectrum, a bifurcation and an equilibrium point analysis. Moreover, its electronic circuit built in the PSpice environment agrees with the numerical results. Besides, in respect of its engineering applications, a novel data compression-encryption scheme based on the new TLN model, matrix factorization theory and compressive sensing technology is introduced for providing a secure data exchange environment in the healthcare community. Finally, performance evaluation indicates that the proposed cryptography scheme has remarkable advantages in terms of reconstruction quality and security.