Chaotic Autonomous Developmental Neural Network
Qiang Lu, Zhaochen Zhang · 2019
The autonomous developmental neural network is proposed based on the investigation on the visual pathways. It integrates both object location and object type from images of learned objects in natural complex backgrounds and is used to recognize and identify the scenes. Chaotic motion can distribute homogeneously within a certain range since it has possessed some uncertainty, ergodicity and stochasticity. Hence, the characteristic of chaotic ergodicity is used to solve the global optimal solution of complex nonlinear problems. In this paper, we propose a chaotic autonomous developmental neural network, in which chaotic sequences are utilized to enhance weighs of neural network. Through simulation and analysis, the results show that the chaotic autonomous developmental neural network can enhance the performance of the basic autonomous developmental neural network.