Learning performance of frequency-modulation digital neural network with on-chip learning

New digital architecture of the frequency-based multilayer neural network (MNN) with on-chip learning is proposed. As the signal level is expressed by the frequency, synaptic multiplier is replaced by a simple frequency converter. Furthermore, the neuron unit using a voting circuit as the nonlinear adder gives better a nonlinear activating function. The backpropagation algorithm is modified for on-chip learning. The proposed MNN architecture was implemented on field programmable gate arrays and various experiments were conducted to test the performance of the system. The experimental results show that the proposed neuron has a very good nonlinear function owing to the voting circuit. The learning behavior of the proposed MNN was also tested by experiments, which show that the proposed MNN has good learning performance and generalization capabilities