Qinru Qiu Professor Syracuse University United States 1 (Northeastern U.S.) Email 2023 2024 Talk(s): A Journey into Neuromorphic Computing: Models, Algorithms, and Implementations A Journey into Neuromorphic Computing: Models, Algorithms, and Implementations × The proliferation of "big data" applications poses significant challenges in terms of speed and scalability for traditional computer systems. The increasing performance gap between CPUs and memory, commonly referred to as the "memory wall," greatly impedes the performance of traditional Von Neumann machines. As a result, neuromorphic computing systems have garnered considerable attention. These systems operate by emulating the charging and discharging processes of neurons and synapse potential in a biologically plausible computing paradigm. Electrical impulses or spikes facilitate inter-neuron communication. The unique encoding of information in the spike domain enables asynchronous event-driven computation and communication, potentially resulting in high energy efficiency. In this seminar, I will introduce several typical computing models of neuron and synapses that can be utilized to build spiking neural networks (SNNs). Additionally, selected inference and learning algorithms for SNNs will be discussed, followed by a brief overview of existing hardware and software solutions for implementing neuromorphic computing. I will further present our Error-Modulated Spike-Timing-Dependent Plasticity (EMSTDP) algorithm, which is capable of supervised training of a deep SNN, and its implementation on a neurosynaptic processor. Compelling results that highlight the potential of this innovative computing paradigm will be presented. Energy Efficient Learning and Adaptation of Multivariate Time Series Using Neuromorphic Computing Energy Efficient Learning and Adaptation of Multivariate Time Series Using Neuromorphic Computing × The dynamics of the physical world can be captured by multi-channel time-varying analog signals. From sensing to actuation, to interact with the physical world, IoTs and edge devices must have the ability to detect, classify, and generate patterns in multivariate time series with rich temporal and spatial dynamics. However, limited hardware resources and battery capacity pose significant challenges in information representation and processing. Additionally, the constantly changing environment and mission requirements demands the ability for online learning and adaptation. Inspired by the structure and behavior of biological neural systems, spiking neural network (SNN) models and neuromorphic computing hardware incorporate many energy-efficient features of biological systems making them effective for mobile and edge applications. The neuron and synapse states maintained by membrane potentials provide rich temporal dynamics for pattern detection and generation, making the model ideal for in-memory computing. In this talk I will introduce SNNs and neuromorphic computing techniques for multivariate time series processing. Using neurons modeled as a network of infinite impulse response filters, the SNN can either work as a classifier to detect patterns in the input temporal sequences or as a generator to generate desired temporal sequences. The ability to discern temporal patterns allows for very sparse input representation, where information is encoded by the intervals between spike events. When combined with event-driven computing and communication, such temporal coding results in significant energy savings.