CAREER: Spiking Neural Circuits and Networks with Temporally Dynamic Learning

职业：具有时间动态学习的尖峰神经回路和网络

• 批准号: 1751230
• 负责人: Kurtis Cantley
• 金额: $ 54.89万
• 依托单位: Boise State University
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2018
• 资助国家: 美国
• 起止时间: 2018-07-01 至 2024-06-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1751230&HistoricalAwards=false
• 关键词: CAREER; Spiking; Neural; Circuits; Networks

The human brain is adept at processing vast amounts of real-world information, learning new concepts, and adapting to changing environments. Emulating the plasticity and cognitive abilities of the brain using electronics will enable new paradigms in computing not currently accessible with any digital system. Achieving this goal requires new, efficient approaches that mimic an array of complex mechanisms observed in biology. Chief among these are the rules governing changes in the strength of connections between neurons. Synaptic connection strengths are believed to be ultimately responsible for memory, reasoning, perception, and other higher-order functions. A combination of established and emerging semiconductor device and circuit technologies will be used in this project to investigate networks with biologically-realistic learning. Integrated education and outreach activities will enhance understanding and awareness of biological and electronic neural networks for middle and high school students, regionally and nationally. This CAREER award supports the development of electronic spiking neural networks (SNNs) that capture the dynamic synaptic learning modalities found in the brain. Circuit blocks with short-term memory will be used in conjunction with devices that change electrical resistance over longer periods to achieve synaptic learning that depends on spiking frequency as well as individual spike times. A primary objective is the design, fabrication, and testing of a neuro-synaptic architecture with learning rates tunable over a range of milliseconds to hours. The physical system will then be used to study cognitive tasks that combine classical conditioning and spatio-temporal pattern recognition. Sensitivity of the network to device variations and control parameters will be examined in terms of pattern recognition accuracy, power consumption, and stability. Benchmarks will also be applied to quantify system performance relative to other techniques. Ideas and concepts generated by this research will ultimately advance the capabilities of intelligent machines employed for a wide variety of tasks such as sensory processing, inference and outcome prediction, anomaly detection, pattern recognition and classification, and big data analytics.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.

人类大脑擅长处理大量的现实世界信息，学习新概念，并适应不断变化的环境。利用电子技术模拟大脑的可塑性和认知能力，将使目前任何数字系统都无法实现的计算新范式成为可能。实现这一目标需要新的，有效的方法，模仿在生物学中观察到的一系列复杂机制。其中最主要的是控制神经元之间连接强度变化的规则。突触连接强度被认为是记忆、推理、感知和其他高阶功能的最终负责者。在这个项目中，将使用现有和新兴的半导体器件和电路技术的组合来研究具有生物现实学习的网络。综合教育和外联活动将提高区域和国家的中学生对生物和电子神经网络的理解和认识。这个CAREER奖支持电子尖峰神经网络（SNN）的开发，该网络捕获大脑中发现的动态突触学习模式。具有短期记忆的电路块将与在较长时间内改变电阻的设备结合使用，以实现取决于尖峰频率和单个尖峰时间的突触学习。主要目标是设计、制造和测试一种神经突触架构，其学习速率可在毫秒到小时的范围内调节。然后，物理系统将被用于研究结合了联合收割机经典条件反射和时空模式识别的认知任务。网络对设备变化和控制参数的敏感性将在模式识别精度、功耗和稳定性方面进行检查。还将采用基准来量化系统相对于其他技术的性能。通过这项研究产生的想法和概念将最终提高智能机器的能力，用于各种各样的任务，如感官处理，推理和结果预测，异常检测，模式识别和分类，以及大数据分析。该奖项反映了NSF的法定使命，并被认为值得通过使用基金会的智力价值和更广泛的影响审查标准进行评估来支持。

项目成果

A Model for R(t) Elements and R(t)-Based Spike-Timing-Dependent Plasticity With Basic Circuit Examples

R(t) 元件模型和基于 R(t) 的尖峰时序相关可塑性以及基本电路示例

• DOI: 10.1109/tnnls.2019.2952768
• 发表时间: 2019
• 期刊: IEEE Transactions on Neural Networks and Learning Systems
• 影响因子: 10.4
• 作者: Ivans, Robert C.;Dahl, Sumedha Gandharava;Cantley, Kurtis D.
• 通讯作者: Cantley, Kurtis D.

A Spatiotemporal Pattern Detector

时空模式检测器

• DOI: 10.1109/mwscas.2019.8884799
• 发表时间: 2019
• 期刊: 2019 IEEE 62nd International Midwest Symposium on Circuits and Systems (MWSCAS
• 作者: Ivans, Robert;Cantley, Kurtis D.
• 通讯作者: Cantley, Kurtis D.