基于余弦函数的神经网络硬件实现

The Design Method of the Circuit on the Cosine Basis Function Neural Networks

研究了正交余弦基函数神经网络模型和学习算法 ,给出了仿真实例 ,并提出了余弦函数基神经网络模型电路实现方案 .仿真结果表明 ,该网络学习收敛速度快 ,可任意逼近非线性映射等优异性能 .此外 ,网络结构简单 ,便于电路实现 ,一个余弦波振荡源经倍频即可实现隐层神经元电路 .

The model,algorithm, emulation examples and hardwired circuit based on cosine basis function neural network are studied in detail. The emulation results show that the network has an execllent performance with its fast astringercy, and it may be arbitrarily imminent to nonlinear function. In addition, its circuit is easy to be designed because of the simple structure of cosine basis function neural network. All the neural cells are constructed with a single cosine oscillator and a series of frequency multiplier.