Abstract:The brain performs intelligent tasks with extremely low energy consumption. This work takes its inspiration from two strategies used by the brain to achieve this energy efficiency: the absence of separation between computing and memory functions and reliance on low-precision computation. The emergence of resistive memory technologies indeed provides an opportunity to tightly co-integrate logic and memory in hardware. In parallel, the recently proposed concept of a Binarized Neural Network, where multiplications are replaced by exclusive NOR (XNOR) logic gates, offers a way to implement artificial intelligence using very low precision computation. In this work, we therefore propose a strategy for implementing low-energy Binarized Neural Networks that employs brain-inspired concepts while retaining the energy benefits of digital electronics. We design, fabricate, and test a memory array, including periphery and sensing circuits, that is optimized for this in-memory computing scheme. Our circuit employs hafnium oxide resistive memory integrated in the back end of line of a 130-nm CMOS process, in a two-transistor, two-resistor cell, which allows the exclusive NOR operations of the neural network to be performed directly within the sense amplifiers. We show, based on extensive electrical measurements, that our design allows a reduction in the number of bit errors on the synaptic weights without the use of formal error-correcting codes. We design a whole system using this memory array. We show on standard machine learning tasks (MNIST, CIFAR-10, ImageNet, and an ECG task) that the system has inherent resilience to bit errors. We evidence that its energy consumption is attractive compared to more standard approaches and that it can use memory devices in regimes where they exhibit particularly low programming energy and high endurance. We conclude the work by discussing how it associates biologically plausible ideas with more traditional digital electronics concepts.

Realization of the conscience mechanism in CMOS implementation of winner-takes-all self-organizing neural networks.

Winner-Takes-All Mechanism Realized By Memristive Neural Network

Efficient In Situ Learning of Hybrid LIF Neurons Using WTA Mechanism for High-Speed Low-Power Neuromorphic Systems

CMOS Implementation of ANNs Based on Analog Optimization of N-Dimensional Objective Functions

Direct CMOS Implementation of Neuromorphic Temporal Neural Networks for Sensory Processing

A New Hybrid Neural System Interfacing Neurons and Silicon Hardware for Fast Signal Recognition

Organizing Sequential Memory in a Neuromorphic Device Using Dynamic Neural Fields

On-chip learning in a conventional silicon MOSFET based Analog Hardware Neural Network

A CMOS Rectified Linear Unit Operating in Weak Inversion for Memristive Neuromorphic Circuits

Exploiting Inherent Error-Resiliency of Neuromorphic Computing to achieve Extreme Energy-Efficiency through Mixed-Signal Neurons

Memristor-based Neuromorphic Implementations for Artificial Neural Networks

A Self-Rectification and Quasi-Linear Analogue Memristor for Artificial Neural Networks

CMOS-based area-and-power-efficient neuron and synapse circuits for time-domain analog spiking neural networks

Digital Biologically Plausible Implementation of Binarized Neural Networks With Differential Hafnium Oxide Resistive Memory Arrays

Versatile CMOS Analog LIF Neuron for Memristor-Integrated Neuromorphic Circuits

Real-Time State Modulation and Acquisition Circuit in Neuromorphic Memristive Systems

CMOS and Memristor-Based Neural Network Design for Position Detection

Synaptic Resistor Circuits Based on Al Oxide and Ti Silicide for Concurrent Learning and Signal Processing in Artificial Intelligence Systems

Self-Activation Neural Network Based on Self-Selective Memory Device with Rectified Multilevel States

A Capacitor-Less CMOS Neuron Circuit for Neuromemristive Networks

Analog Spiking Neuron in CMOS 28 nm Towards Large-Scale Neuromorphic Processors