Abstract:Deep neural networks are the most in demand and popular for solving a wide range of problems in various fields of science and technology due to their learning ability. Increasing the number of layers in software implementation of deep neural networks (DNN) requires increased power consumption when working with big asynchronous data and becomes ineffective. To solve this problem, this study proposes a new approach to DNN design with on-chip training, based on the analog implementation of all computational operations, including matrix–vector multiplication in synaptic crossbar arrays, with digital implementation of synaptic weight setup and storage operations. The efficiency of the proposed approach is determined by the developed integrated synaptic CMOS IP block, which performs direct (without DAC) linear conversion of the binary code of the synaptic weight into the level of synaptic conductivity, as well as unified analog CMOS IP blocks with the ability to change the number of inputs and outputs. The paper presents the circuit diagrams, topology and results of SPICE modeling of CMOS IP blocks, as well as a description of the functional capabilities of the compiler, which allows using these IP blocks to synthesize functional circuits and topology of deep neural networks with specified parameters based on a neural network graph with training using the backpropagation method.

Analogue Circuit Realization of a Programmable Sigmoidal Function and Its Derivative for On-Chip BP Learning

A programmable on-chip BP learning neural network with enhanced neuron characteristics

Hardware implementation of an on-chip BP learning neural network with programmable neuron characteristics and learning rate adaptation

An On-Chip BP Learning Neural Network with Ideal Neuron Characteristics and Learning Rate Adaptation

Circuit Design of an Adjustable Neuron Activation Function and Its Derivative

Method for implementing feedforward neural network by using digital and analog circuits

Programmable synapse analog CMOS neural network

Hardware implementation of single-bit nonlinear BP neural network based on sigma-delta modulation

Hardware Realization of Building Blocks for Artificial Neural Networks

Implementation of activated function of neural networks by using hybrid CORDIC

Adaptive Integratable Hardware Realization of Analog Neural Networks for Nonlinear System

Hardware Implementation of On-Chip Learning Neural Network on FPGA

Design of a new CMOS Low-Power Analogue Neuron

Automated design of deep neural networks with in-situ training architecture based on analog functional blocks

Global multistability and analog circuit implementation of an adapting synapse-based neuron model

Synthetical backpropagation algorithm

A Design and Implementation of Reconfigurable Architecture for Neural Networks Based on Systolic Arrays

Circuit Implementation of Proximal Projection Neural Networks for Composite Optimization Problems.

Sign backpropagation: An on-chip learning algorithm for analog RRAM neuromorphic computing systems

Full-Analog Implementation of Activation Function Based on Phase-Change Memory for Artificial Neural Networks

In-Hardware Learning of Multilayer Spiking Neural Networks on a Neuromorphic Processor