Abstract:In recent years, the scaling down that Moore’s Law relies on has been gradually slowing down, and the traditional von Neumann architecture has been limiting the improvement of computing power. Thus, neuromorphic in-memory computing hardware has been proposed and is becoming a promising alternative. However, there is still a long way to make it possible, and one of the problems is to provide an efficient, reliable, and achievable neural network for hardware implementation. In this paper, we proposed a two-layer fully connected spiking neural network based on binary MRAM (Magneto-resistive Random Access Memory) synapses with low hardware cost. First, the network used an array of multiple binary MRAM cells to store multi-bit fixed-point weight values. This helps to simplify the read/write circuit. Second, we used different kinds of spike encoders that ensure the sparsity of input spikes, to reduce the complexity of peripheral circuits, such as sense amplifiers. Third, we designed a single-step learning rule, which fit well with the fixed-point binary weights. Fourth, we replaced the traditional exponential Leak-Integrate-Fire (LIF) neuron model to avoid the massive cost of exponential circuits. The simulation results showed that, compared to other similar works, our SNN with 1184 neurons and 313,600 synapses achieved an accuracy of up to 90.6% in the MNIST recognition task with full-resolution (28 × 28) and full-bit-depth (8-bit) images. In the case of low-resolution (16 × 16) and black-white (1-bit) images, the smaller version of our network with 384 neurons and 32,768 synapses still maintained an accuracy of about 77%, extending its application to ultra-low-cost situations. Both versions need less than 30,000 samples to reach convergence, which is a >50% reduction compared to other similar networks. As for robustness, it is immune to the fluctuation of MRAM cell resistance.

Dropout neuronal unit with tunable probability based on NbO

Dropout neuronal unit with tunable probability based on NbO<sub>x</sub> stochastic memristor for efficient suppression of overfitting

One Memristor-One Electrolyte-Gated Transistor based High Energy-Efficient Dropout Neuronal Units

Implementation of Dropout Neuronal Units Based on Stochastic Memristive Devices in Neural Networks with High Classification Accuracy

Unlocking the Non-deterministic Computing Power with Memory-Elastic Multi-Exit Neural Networks

Adaptive Dropout Method Based on Biological Principles

Rethinking the Usage of Batch Normalization and Dropout in the Training of Deep Neural Networks

Adaptive Sparse Dropout: Learning the Certainty and Uncertainty in Deep Neural Networks

Continuous Dropout

Spatial-SpinDrop: Spatial Dropout-based Binary Bayesian Neural Network with Spintronics Implementation

Design of NbOx memristive neuron and its application in spiking neural networks

Advanced Dropout: A Model-free Methodology for Bayesian Dropout Optimization.

Mitigate IR-Drop Effect by Modulating Neuron Activation Functions for Implementing Neural Networks on Memristor Crossbar Arrays

Continuously evolving dropout with multi-objective evolutionary optimisation

A Low-Cost Hardware-Friendly Spiking Neural Network Based on Binary MRAM Synapses, Accelerated Using In-Memory Computing

A noise-tolerant, resource-saving probabilistic binary neural network implemented by the SOT-MRAM compute-in-memory system

Normalization and Dropout for Stochastic Computing-Based Deep Convolutional Neural Networks.

Approximate Random Dropout.

A Memristor-Based Spiking Neural Network With High Scalability and Learning Efficiency

SOT-MRAM-Enabled Probabilistic Binary Neural Networks for Noise-Tolerant and Fast Training