Energy-efficient stochastic computing with superparamagnetic tunnel junctions

TL;DR

This paper introduces an energy-efficient stochastic computing method using superparamagnetic tunnel junctions (SMTJs) and a novel neural network architecture, achieving significant energy savings and high accuracy on MNIST.

Contribution

It presents a digitally programmable, energy-efficient bitstream generator based on SMTJs and integrates it into a co-designed neural network architecture for improved energy efficiency.

Findings

Achieves approximately 150 nJ per inference on MNIST.

Provides 97% accuracy on MNIST dataset.

Offers 1.4 to 7.7 times better energy efficiency than recent methods.

Abstract

Superparamagnetic tunnel junctions (SMTJs) have emerged as a competitive, realistic nanotechnology to support novel forms of stochastic computation in CMOS-compatible platforms. One of their applications is to generate random bitstreams suitable for use in stochastic computing implementations. We describe a method for digitally programmable bitstream generation based on pre-charge sense amplifiers. This generator is significantly more energy efficient than SMTJ-based bitstream generators that tune probabilities with spin currents and a factor of two more efficient than related CMOS-based implementations. The true randomness of this bitstream generator allows us to use them as the fundamental units of a novel neural network architecture. To take advantage of the potential savings, we codesign the algorithm with the circuit, rather than directly transcribing a classical neural network into hardware. The flexibility of the neural network mathematics allows us to adapt the network to the explicitly energy efficient choices we make at the device level. The result is a convolutional neural network design operating at $\approx$ 150 nJ per inference with 97 % performance on MNIST -- a factor of 1.4 to 7.7 improvement in energy efficiency over comparable proposals in the recent literature.