Memory resistor neural network training method based on fuzzy Boltzmann machine

Abstract

The invention discloses a memory resistor neural network training method based on a fuzzy Boltzmann machine, and the method comprises the steps: employing a fuzzification processing method, changing the connection intensity / weight value in a limited Boltzmann machine network to a fuzzy number from a definite number, and obtaining a fuzzy weight value; substituting the fuzzy weight value into a limited Boltzmann machine, and obtaining a fuzzy limited Boltzmann machine network which is more suitable for the description of the characteristics of a memory resistor device, wherein the network training process is the updating of the fuzzy weight value, thereby obtaining a trained memory resistor neural network. The method avoids the impact on the network precision and stability from the fluctuation of a device when a memory resistor is taken as a synapse unit in neural network hardware, can improve the robustness of the neural network learning, is universal, and can serve as a universal method for building a neural form system with the inherent random fluctuation for a processing device.

Description

technical field [0001] The invention belongs to the field of information technology and relates to neural network computing technology, in particular to a memristive neural network training method based on a fuzzy Boltzmann machine. Background technique [0002] Neuromorphic computing provides a new way to overcome the limitations of traditional von Neumann computer architecture in terms of efficiency and speed, and proposes a new computing architecture suitable for large-scale parallel computing and high energy efficiency. Memristors, as the fourth passive circuit element, are considered to be perfect candidates for electronic synaptic structures in neuromorphic systems due to their own properties similar to synaptic structures in neural networks. On the one hand, brain-like computing algorithms currently rely heavily on the multiplication of vectors to compute the output of neurons in the network, and high-density memristor crossbar arrays naturally provide the ability to ...

AI Technical Summary

Problems solved by technology

[0004] In order to overcome the deficiencies in the prior art above, the present invention provides a memristive neural network training method based on a fuzzy Boltzmann machine, which can solve the problem of random fluctuations and fluctuations in electronic synapses in the current memristor-based memristive neural network. Issues Affecting Neural Network Performance

Images