Hardware Implementation of Neural Networks with Learning Capability Using Simultaneous Perturbation

点击翻译按钮获取中文摘要

In this research, we demonstrated feasibility of a learning rule using the simultaneous perturbation optimization method for artificial neural networks. First, we showed that combination of pulse density expression and the simultaneous perturbation method is useful for hardware implementation of neural networks. Second, we fabricated support vector machine with learning mechanism using the simultaneous perturbation method based on FPGA. Finally, a pulse coupled oscillator with learning capability is realized as an analog circuit system using FPAA and the simultaneous perturbation method.