A Memristor-Based Randomly Dropping Neurons

Abstract

The invention belongs to the technical field of semiconductor information, and discloses a memristor-based randomly discarding neurons. The random discarding neuron is used to choose the received signal, and a discarding control unit and a MOS switch are arranged in it, and the discarding control unit is used to convert the received signal into a control instruction, thereby controlling the switch of the MOS switch, which includes a memristor The memristor is used to receive the excitation signal and convert it into a random current signal, the voltage divider resistor is used to convert the random current signal into a random voltage signal, and the comparator is used to compare the random voltage signal with the preset The threshold is compared, and the register is used to accept the signal from the comparator and pass it to the MOS switch as a control command, thereby controlling the switch of the MOS. Through the invention, the recognition accuracy of the neural network is improved, the requirement for hardware artificial synapses is reduced, and the problems of over-fitting and nonlinear weight update in the training process of the neural network are solved.

Description

technical field [0001] The invention belongs to the technical field related to semiconductor information, and more specifically relates to a method of randomly discarding neurons based on a memristor. Background technique [0002] With the great success of software-based artificial neural networks, in order to further improve the efficiency of neural networks, research on hardware accelerators for artificial neural networks has been started to achieve the purpose of increasing computing efficiency and reducing computing power consumption. [0003] At this stage, those skilled in the art have done some research, such as using a memristor with a simple sandwich structure to simulate most of the functions of synapses, and showing the continuous conductance of memristors under the action of artificial neurons. Increase or decrease, and realize the simulation of some important learning rules of the nervous system. However, due to the continuous progress of the algorithm, more re...

AI Technical Summary

Problems solved by technology

[0004] For the above defects or improvement needs of the prior art, the present invention provides a randomly discarding neuron based on a memristor, wherein the design of the structure and function of its key component discarding control unit makes the discarding neuron in the present invention At the same time, some neurons are randomly discarded, the recognition accuracy of the neural network is improved, and the requirements for hardware artificial synapses are reduced, thereby solving the over-fitting problem in the training process of the neural network and the problem of nonlinear weight update of the hardware artificial synapses. question

Images