E2CDA: Type I: Probabilistic Spin Logic for Low-Energy Boolean and Non-Boolean Computing

E2CDA：类型 I：用于低能量布尔和非布尔计算的概率自旋逻辑

• 批准号: 1739635
• 负责人: Joerg Appenzeller
• 金额: $ 245.38万
• 依托单位: Purdue University
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2017
• 资助国家: 美国
• 起止时间: 2017-10-01 至 2021-09-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1739635&HistoricalAwards=false
• 关键词: E2CDA; Type; Probabilistic; Spin; Logic

Various components of hardware and software have come together to make this electronic revolution surrounding smartphones and the like possible. To further expand our capabilities, to enter a new era of even more comprehensive networking between individual users and the cloud, as well as to solve problems of a complexity level that cannot be tackled even with today's computers, there is the need to push the frontiers of data processing beyond anything that can be achieved through "conventional" electronics. One approach is to improve the hardware further. However the latter has been historically mainly focused on making devices ever smaller, thus gaining the benefit of a more compact but very power hungry electronics. The current project revolves around a novel device that uses electron spins rather than electron charges for information processing, and holds the promise to achieve the above mentioned goals at a power consumption level that is far smaller than what can be envisioned with conventional technologies. At the same time the project aims at a new level of education and dissemination through the concept that we call "atoms to systems". A new generation of engineers requires skill sets and a knowledge base that is different from what was learned previously. A high priority of the project is thus to make the models and ideas that underlie them accessible across science and engineering.There is increasing interest in a fundamentally different form of brain-like logic based on probabilistic inference that is far more effective and energy efficient in dealing with the problems of search and recognition posed by the ever increasing amounts of "big data". Probabilistic logic is currently implemented with software algorithms that run on a deterministic computing platform. The goal is to lay the foundation for a new P(robabilistic)-computing platform using unstable multiferroics, a manmade material combination that combines distinct electrical, mechanical and magnetic properties. Unlike quantum computers, P-computers should operate robustly at room temperature, while providing some aspects of the "quantum parallelism" that facilitates the solution of hard problems. The team will work on the material, device and circuit development of this novel probabilistic computing idea. Technical work related to the characterization of novel material properties, the impact of various structural parameters and architectural aspects will all be explored in parallel by the group of experts.

硬件和软件的各种组成部分都聚集在一起，使围绕智能手机等的电子革命成为可能。为了进一步扩大我们的能力，进入个人用户和云之间更全面的网络的新时代，以及解决即使在当今计算机的情况下也无法解决的复杂性问题，需要将数据处理的边界推广到可以通过“常规”电子产品来实现的任何事情。一种方法是进一步改善硬件。然而，后者历史上一直主要集中于使设备变得越来越小，从而获得了更紧凑但非常饥饿的电子设备的好处。当前的项目围绕着使用电子旋转而不是电子费用进行信息处理的新型设备展开，并承诺在功耗水平上实现上述目标，远小于传统技术所能设想的目标。同时，该项目通过我们称为“原子到系统”的概念来实现新的教育和传播水平。新一代的工程师需要技能集和与以前学到的不同的知识基础。因此，该项目的一个很高的优先事项是使这些模型和思想是基于它们在科学和工程之间访问的模型和思想。基于概率推断，人们对从根本不同的类似脑的逻辑形式的逻辑形式越来越有兴趣，而概率的推论则更有效，更有效地应对搜索和认可的问题，而越来越多的“大数据”。概率逻辑当前是通过在确定性计算平台上运行的软件算法实现的。目的是为新的P（robabilistic）使用不稳定的多效率（一种人造的材料组合）奠定基础，该平台结合了不同的电气，机械和磁性。与量子计算机不同，P计算机应在室温下稳健地运行，同时提供“量子并行性”的某些方面，以促进解决硬问题的解决方案。该团队将研究这个新型概率计算想法的材料，设备和电路开发。与新型材料特性的特征相关的技术工作，各种结构参数和建筑方面的影响都将由专家组并行探索。

项目成果

ApGAN: Approximate GAN for Robust Low Energy Learning From Imprecise Components

• DOI: 10.1109/tc.2019.2949042
• 发表时间: 2020-03-01
• 期刊: IEEE TRANSACTIONS ON COMPUTERS
• 影响因子: 3.7
• 作者: Roohi, Arman;Sheikhfaal, Shadi;DeMara, Ronald F.
• 通讯作者: DeMara, Ronald F.

Hybrid spin-CMOS stochastic spiking neuron for high-speed emulation of In vivo neuron dynamics

• DOI: 10.1049/iet-cdt.2017.0145
• 发表时间: 2018-02
• 期刊: IET Comput. Digit. Tech.
• 作者: Steven D. Pyle;Kerem Y Çamsarı;R. Demara

Composable Probabilistic Inference Networks Using MRAM-based Stochastic Neurons

使用基于 MRAM 的随机神经元的可组合概率推理网络

• 发表时间: 2019
• 期刊: ACM journal on emerging technologies in computing systems
• 影响因子: 2.2
• 作者: Zand, Ramtin;Camsari, kerem Y.;Datta, Supriyo;DeMara, Ronald F.
• 通讯作者: DeMara, Ronald F.

Leveraging Stochasticity for In Situ Learning in Binarized Deep Neural Networks

利用随机性进行二值化深度神经网络的原位学习

• DOI: 10.1109/mc.2019.2906133
• 发表时间: 2019
• 期刊: Computer
• 影响因子: 2.2
• 作者: Pyle, Steven D.;Sapp, Justin D.;DeMara, Ronald F.
• 通讯作者: DeMara, Ronald F.

A Novel Compound Synapse Using Probabilistic Spin–Orbit-Torque Switching for MTJ-Based Deep Neural Networks

• DOI: 10.1109/jxcdc.2019.2956468
• 发表时间: 2019-10
• 期刊: IEEE Journal on Exploratory Solid-State Computational Devices and Circuits
• 影响因子: 2.4
• 作者: V. Ostwal;Ramtin Zand;R. Demara;J. Appenzeller