[논문 리뷰] A Survey of Neuromorphic Computing and Neural Networks in Hardware

Catherine D. Schuman, Thomas E. Potok외 5명|arXiv (Cornell University)|2017. 05. 19.

Advanced Memory and Neural Computing참고 문헌 1,597인용 수 362

한 줄 요약

neuromorphic computing 및 neural network hardware에 대한 포괄적 35-year 간 조사로, 동기, 모델, 학습 알고리즘, 하드웨어 구현, 지원 시스템 및 응용 분야를 상세히 다루고, 향후 연구 격차를 확인한다.

ABSTRACT

Neuromorphic computing has come to refer to a variety of brain-inspired computers, devices, and models that contrast the pervasive von Neumann computer architecture. This biologically inspired approach has created highly connected synthetic neurons and synapses that can be used to model neuroscience theories as well as solve challenging machine learning problems. The promise of the technology is to create a brain-like ability to learn and adapt, but the technical challenges are significant, starting with an accurate neuroscience model of how the brain works, to finding materials and engineering breakthroughs to build devices to support these models, to creating a programming framework so the systems can learn, to creating applications with brain-like capabilities. In this work, we provide a comprehensive survey of the research and motivations for neuromorphic computing over its history. We begin with a 35-year review of the motivations and drivers of neuromorphic computing, then look at the major research areas of the field, which we define as neuro-inspired models, algorithms and learning approaches, hardware and devices, supporting systems, and finally applications. We conclude with a broad discussion on the major research topics that need to be addressed in the coming years to see the promise of neuromorphic computing fulfilled. The goals of this work are to provide an exhaustive review of the research conducted in neuromorphic computing since the inception of the term, and to motivate further work by illuminating gaps in the field where new research is needed.

연구 동기 및 목표

• neuromorphic computing 분야에 대한 역사적이고 폭넓은 개요를 제공한다.
• 시간에 따라 neuromorphic hardware 개발을 이끄는 동기를 요약한다.
• 하드웨어 구현에 사용되는 neuron, synapse, 및 network 모델를 나열한다.
• neuromorphic 시스템에 적용 가능한 학습 알고리즘과 학습 방법을 검토한다.
• 하드웨어 구현, 지원 시스템 및 응용 분야를 조사하고, 열려 있는 연구 질문을 식별한다.

제안 방법

• 3,000 papers에 걸친 35년 간의 조사.
• 작업을 모델, 알고리즘/학습, 하드웨어, 지원 시스템, 응용 분야로 분류한다.
• 주요 neuron 및 synapse 모델과 하드웨어 구현에 대한 논의를 다룬다.
• 저전력, 실시간 성능, 확장성과 같은 목표에 대한 분석.
• 격차 및 향후 연구 방향의 식별.

실험 결과

연구 질문

• RQ1 What historical motivations have shaped neuromorphic computing and how have they evolved?
• RQ2 What neuron, synapse, and network models have hardware realizations, and how do they differ?
• RQ3 What learning rules and algorithms are prevalent in neuromorphic hardware, and how are they implemented?
• RQ4 What hardware architectures and materials are used, and what applications have they enabled?
• RQ5 What are the major open challenges and gaps for advancing neuromorphic computing?

주요 결과

• Interest in neuromorphic computing has risen strongly in the last decade.
• Low-power operation is the dominant motivation for modern neuromorphic hardware.
• A wide variety of neuron models exist, spanning biologically plausible, biologically inspired, integrate-and-fire, and McCulloch-Pitts traditions.
• Synapse models often emphasize plasticity and STDP, with diverse synaptic dynamics and network learning rules.
• Hardware implementations span spiking and non-spiking networks, with different device technologies and programming capabilities.
• The field remains multi-disciplinary, addressing materials science, neuroscience, device engineering, and algorithm/software development, with clear open research gaps.