Toward a general-purpose analog VLSI neural network with on-chip learning

Abstract

This paper describes elements necessary for a general-purpose low-cost very large scale integration (VLSI) neural network. By choosing a learning algorithm that is tolerant of analog nonidealities, the promise of high-density analog VLSI is realized. A 64-synapse, 8-neuron proof-of-concept chip is described. The synapse, which occupies only 4900 mum(2) in a 2-mum technology, includes a hybrid of nonvolatile and dynamic weight storage that provides fast and accurate learning as well as reliable long-term storage with no refreshing. The architecture is user-configurable in any one-hidden-layer topology. The user-interface is fully microprocessor compatible. Learning is accomplished with minimal external support; the user need only present inputs, targets, and a clock. Learning is fast and reliable. The chip solves four-bit parity in an average of 680 ms and is successful in about 96% of the trials.

Uso de um Ambiente Codesign para a Implementação de Redes Neurais

“…Para a implementação em hardware dedicado pode-se abordar a forma do projeto sob três focos distintos. O primeiro foco utiliza técnicas digitais de implementação, [2], o segundo foco aborda as técnicas analógicas, como pode ser examinado em [3], [4], [5], [6], [7] e [8]; e por fim, o terceiro foco se utiliza das técnicas híbridas, como em [9] e em [10], tendo a implementação composta em parte digital e em parte analógica.…”

Deviation-tolerant floating gate structures as a way to design an on-chip learning neural networks

“…9. The source of increment transistor is connected to V dd during change of the weight and to the ground during holding the weight [9]. Similarly decrement transistor is connected to the ground and V dd in the opposite way.…”

A feed-forward time-multiplexed neural network with mixed-signal neuron–synapse arrays

“…Analog neural networks employing analog circuitry provide power-and area-efficient implementations of the required processing functions such as summation, multiplication, and sigmoid function [1][2][3][4]. Regardless of their efficiency analog networks suffer from offsets, mismatches and limited dynamic range and accuracy.…”