Development of a low-power deep-learning chip using adiabatic superconducting technology

2020 Fiscal Year Final Research Report

• Project/Area Number: 19K15041
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 21060:Electron device and electronic equipment-related
• Research Institution: Yokohama National University
• Principal Investigator: Chen Olivia 横浜国立大学, 先端科学高等研究院, 特任教員(助教) (70837856)
• Project Period (FY): 2019-04-01 – 2021-03-31
• Keywords: 電子工学 / 超伝導エレクトロニクス / 集積回路工学

Outline of Final Research Achievements

In this research, we proposed an extremely low-power deep learning chip by combining Stochastic Computing (SC), an approximate computing scheme with the adiabatic quantum magnetic flux parametron (AQFP), a low-power superconducting technology. As the research results, we have developed an automated design tool chain for large-scale superconducting circuits design, prototyped the proposed deep learning chip, and demonstrated its operation at low temperatures.

Free Research Field

電子デバイス・電子機器

Academic Significance and Societal Importance of the Research Achievements

人工知能の急速な発展に伴う情報量が爆発的増大し，莫大な電力が消費されてしまいます．本研究成果では，半導体回路に対して５桁以上消費電力効率を持つ超伝導回路を基盤技術とする上，近似計算である新たな計算方式との連携を通じて，1Wで千兆回演算級のエネルギー効率を有する新たな超低電力AIシステムの開発に挑戦します．また，本技術の応用の拡大，引いては地球温暖化の防止にもつながると考えます．