I-Corps: Neuromorphic device derived from resistive switching system

这个i-Corps项目的更广泛的影响/商业潜力是促进各种行业的创新，这些行业致力于先进的基于计算的应用，这是可靠的神经形态芯片(神经网络被蚀刻在硅中)的结果。这项技术可能会加速机器学习的进步，并通过克服冯·诺伊曼瓶颈最终实现计算体系结构的突破--冯·诺伊曼瓶颈的内在延迟源于处理器和内存的分离。在该项目技术的神经形态特征的支持下，机器学习可能在设计和能源消耗方面以前所未有的速度和效率进行。因此，基于密集机器学习的技术，如人脸识别、自动驾驶和其他人工智能领域，将大幅进步。该项目的成功执行还将促进利用人工智能和/或神经形态计算的新兴应用的广泛商业化。这个i-Corps项目利用了最近与高度可控的神经形态设备相关的创新研究。这项技术基于基于氧化物的电阻随机存取存储器(ReRAM)，这是一种基于电刺激诱导的电阻变化的非易失性存储器，主要通过每个电池中所谓的纳米级导电丝的形成和断裂来实现。ReRAM细胞与神经形态系统的内在相似性引发了最近相关研究的热潮。然而，仍然存在相当大的技术问题，包括操作电压、电池电流和操作速度在电池之间的广泛变化。对物理交换机制的不清楚理解排除了为商业设备设计可靠单元的基本原理。拟议的新计划，其中的不可控性被显著降低，预计将在实现可商业化的神经形态设备方面实现突破。这一奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

The broader impact/commercial potential of this I-Corps project is in facilitating innovations in a wide range of industries working on advanced computing-based applications as the result of a reliable neuromorphic chip (where neural networks are etched into silicon). This technology may accelerate the advance of machine learning and ultimately realize a breakthrough in computing architecture by overcoming the Von Neumann bottleneck -- where the intrinsic latency originated from the separation of processor and memory. Empowered by neuromorphic features of this project's technology, machine learning may be carried out with unprecedented speed and efficiency in design and energy consumption. Hence, intensive machine learning based technologies such as face recognition, autonomous driving and other areas of artificial intelligence will be advanced by a significant margin. A successful execution of the project will also facilitate widespread commercialization of emerging applications utilizing artificial intelligence and/or neuromorphic computing. This I-Corps project leverages the recent innovative research related to highly controllable neuromorphic devices. The technology is based on the oxide-based resistive random access memory (ReRAM), a non-volatile memory based upon electrical stimuli-induced resistance changes, mostly by the formation and rupture of so-called nanoscale conducting filaments in each cell. The intrinsic resemblance of ReRAM cells to neuromorphic systems triggered a recent boom in related research. However, there are still considerable technical issues including wide cell-to-cell variations in operation voltage, cell current and operation speed. Unclear understanding of physical switching mechanisms precludes a rationale design of reliable cells for a commercial device. The proposed novel scheme, in which the uncontrollability is significantly minimized, is expected to achieve a breakthrough in realizing commercializable neuromorphic devices.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.