Adiabatic Quantum Computation Applied to Deep Learning Networks.

Jeremy Liu,Thomas Potok,Steven Young,Gangotree Chamka,Garrett Rose,Robert Patton,Ke-Thia Yao,Catherine Schuman,Federico Spedalieri

doi:10.3390/e20050380

Abstract

Training deep learning networks is a difficult task due to computational complexity, and this is traditionally handled by simplifying network topology to enable parallel computation on graphical processing units (GPUs). However, the emergence of quantum devices allows reconsideration of complex topologies. We illustrate a particular network topology that can be trained to classify MNIST data (an image dataset of handwritten digits) and neutrino detection data using a restricted form of adiabatic quantum computation known as quantum annealing performed by a D-Wave processor. We provide a brief description of the hardware and how it solves Ising models, how we translate our data into the corresponding Ising models, and how we use available expanded topology options to explore potential performance improvements. Although we focus on the application of quantum annealing in this article, the work discussed here is just one of three approaches we explored as part of a larger project that considers alternative means for training deep learning networks. The other approaches involve using a high performance computing (HPC) environment to automatically find network topologies with good performance and using neuromorphic computing to find a low-power solution for training deep learning networks. Our results show that our quantum approach can find good network parameters in a reasonable time despite increased network topology complexity; that HPC can find good parameters for traditional, simplified network topologies; and that neuromorphic computers can use low power memristive hardware to represent complex topologies and parameters derived from other architecture choices.

Highlights

A neural network is a machine learning concept originally inspired by studies of the visual cortex of the brain
Before implementing the restricted Boltzmann machines (RBMs) running on Modified National Institute of Standards and Technology (MNIST) data we wanted to get initial results indicating there was some merit to the LBM topology
300,000 cores and 18,000 Nvidia Tesla K20x graphical processing units (GPUs). These results demonstrated that near optimal hyperparameters for convolutional neural networks (CNNs) architectures can be found for the MNIST handwritten digit dataset by combining evolutionary algorithms and high performance computing

Summary

Introduction

A neural network is a machine learning concept originally inspired by studies of the visual cortex of the brain. Neural networks are the neurons of the brain connected to each other via synapses; in machine learning, they are graphical models where variables are connected to each other with certain weights. Both are highly useful in analyzing image data, but practical considerations regarding network topology limit the potential of simulating neural networks on computers. Rather than explicitly choosing one solution or another, these approaches are meant to augment each other Describing these different approaches necessitates a brief description of various machine learning models and networks including Boltzmann machines (BMs), convolutional neural networks (CNNs), and spiking neural networks (SNNs)

Objectives

Results

Discussion

Conclusion