On-Chip학습기능을 가진 확률연산 펄스형 디지털 신경망의 구현

초록

In this paper, RBF(Radial Basis Function) Neural Network is suggested for the prediction of a nonlinear system response. RBF is the model which is based on the locally tuned responses of biological neurons. This model has very fast learning time and is capable of predicting a system behavior with noisy or distorted information. Suggested RBF neural network proposes a way to choose parameters which determine the most suitable RBF to use before learning, whereas with existing RBF neural networks, parameters are determined heuristically. Another key feature of this model is to calculate apIn this paper, an on-chip learning pulse-mode digital neural network with a massively parallel yet compact and flexible network architecture is suggested. Algebraic neural operations are replaced by stochastic processes using pseudo-random sequences and simple logic gates are used as basic computing elements. Using Back-propagation algorithm both feed-forward and learning phases are efficiently implemented with simple logical gates. RNG architecture using LFSR and barrel shifter are adopted to avoid some correlation between pulse trains. Suggested network is designed in digital circuit and its performance is verified by computer simulation.propriate total number of RBF's in hidden layer by forward selection method.