Market assessment of an Spiking Neuron Implementation in Digital Hardware using a Sampling-Based Approach for Reduced Power Consumption

使用基于采样的方法降低功耗，对数字硬件中的尖峰神经元实现进行市场评估

• 批准号: 576556-2022
• 负责人: Mirhassani, MitraM
• 金额: $ 1.09万
• 依托单位: University of Windsor
• 依托单位国家: 加拿大
• 项目类别: Idea to Innovation
• 财政年份: 2022
• 资助国家: 加拿大
• 起止时间: 2022-01-01 至 2023-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=751907
• 关键词: Market; assessment; Spiking; Neuron; Implementation

Vast amounts of data are being generated and processed every single minute by AI solutions. Speed and power are becoming increasingly important factors for companies to consider and many have been looking toward neuromorphic systems as the answer. Neuromorphic systems model elements of a computer after systems in the human brain and nervous system. One of the most characteristic properties of real neural systems is their remarkably low power consumption in comparison to electrical hardware systems. Given the large networks of parallel neurons found in real biological systems, low power consumption is paramount to an effective and feasible electrical hardware replication. Many proposed spiking neuron and neuro-processor implementations offer substantial power savings compared to implementations prior to their proposal, however in each case limitations or restrictions are imposed on the implementation such as the use of a specific technology node, neuron model, and/or hardware implementation technique are required. The proposed patent-pending novel approach to digital hardware implementations of spiking neurons evaluates the neuron's differential equations at a frequency dependent on input current using a sampling-based approach. The variability in the frequency of the sampling reduces unnecessary switching activity for low-stimulus states. This novel Input-DEpendent Variable Sampling (I-DEVS) digital realization method to spiking neuron implementation results in neurons with minimal additional hardware resource usage in exchange for dramatic dynamic power savings as the switching activity is greatly reduced compared to traditional neuron implementations. This proposal will be working on a market assessment to better understand how to bring this innovative technology from the lab to the market.

人工智能解决方案每分钟都会生成和处理大量数据。速度和功率正成为公司需要考虑的越来越重要的因素，许多公司一直在寻求神经形态系统作为答案。神经形态系统模仿人脑和神经系统中的系统来模拟计算机的元素。真实神经系统最典型的特性之一是与电气硬件系统相比，其功耗非常低。考虑到真实生物系统中存在的大型并行神经元网络，低功耗对于有效且可行的电气硬件复制至关重要。许多提出的尖峰神经元和神经处理器实现与提出之前的实现相比提供了大量的功耗，但是在每种情况下对实现都施加了限制或约束，例如使用特定的技术节点、神经元模型和/或硬件需要实施技术。所提出的正在申请专利的尖峰神经元数字硬件实现新方法使用基于采样的方法以取决于输入电流的频率评估神经元的微分方程。采样频率的可变性减少了低刺激状态的不必要的切换活动。这种新颖的输入相关变量采样 (I-DEVS) 数字实现方法可实现尖峰神经元，从而使神经元只需使用最少的额外硬件资源即可换取显着的动态功耗，因为与传统神经元实现相比，开关活动大大减少。该提案将进行市场评估，以更好地了解如何将这项创新技术从实验室推向市场。