Advances in Neuromorphic Computing and Spiking Neural Networks

The recent developments in the field of neuromorphic computing and spiking neural networks (SNNs) have shown significant advancements in energy efficiency, computational speed, and integration with emerging hardware technologies. Researchers are increasingly focusing on leveraging the unique properties of SNNs, such as event-driven processing and temporal encoding, to design systems that mimic biological neural networks for more efficient computation. Key innovations include the development of novel neuron models, efficient training algorithms, and hardware accelerators that integrate SNNs with technologies like photonic computing and resistive random-access memory (ReRAM). These advancements are paving the way for ultra-low-power edge computing applications, particularly in areas like object detection, semantic segmentation, and speech command recognition. Notable contributions include the introduction of compact high-voltage functional electrical stimulation devices, ultra-low power integrate-and-fire neuron circuits, and efficient event-based semantic segmentation networks. These developments not only enhance the performance of SNNs but also make them more viable for real-world applications by addressing challenges related to power consumption, latency, and scalability.

Noteworthy Papers:

Neuro-Photonix: Introduces a near-sensor neuro-symbolic AI computing accelerator on silicon photonics, significantly reducing energy consumption and latency.
CREST: Proposes a spike-driven framework for event-based object detection, achieving superior performance and energy efficiency.
Spike2Former: Enhances SNN performance in image segmentation with a novel architecture and normalized integer spiking neurons.

Sources

Neuro-Photonix: Enabling Near-Sensor Neuro-Symbolic AI Computing on Silicon Photonics Substrate

Sampling from exponential distributions in the time domain with superparamagnetic tunnel junctions

The Reliability Issue in ReRam-based CIM Architecture for SNN: A Survey

SparseMap: Loop Mapping for Sparse CNNs on Streaming Coarse-grained Reconfigurable Array

Deployment Pipeline from Rockpool to Xylo for Edge Computing

Optimal Gradient Checkpointing for Sparse and Recurrent Architectures using Off-Chip Memory

Development of a Compact High-Voltage Functional Electrical Stimulation Device

Design and Performance Analysis of an Ultra-Low Power Integrate-and-Fire Neuron Circuit Using Nanoscale Side-contacted Field Effect Diode Technology

Investigating the Effect of Electrical and Thermal Transport Properties on Oxide-Based Memristors Performance and Reliability

CREST: An Efficient Conjointly-trained Spike-driven Framework for Event-based Object Detection Exploiting Spatiotemporal Dynamics

Noise-based Local Learning using Stochastic Magnetic Tunnel Junctions

Efficient Event-based Semantic Segmentation with Spike-driven Lightweight Transformer-based Networks

Efficient Speech Command Recognition Leveraging Spiking Neural Network and Curriculum Learning-based Knowledge Distillation

Stochastic Analysis of Retention Time of Coupled Memory Topology

An introduction to reservoir computing

Combining Aggregated Attention and Transformer Architecture for Accurate and Efficient Performance of Spiking Neural Networks

Faster and Stronger: When ANN-SNN Conversion Meets Parallel Spiking Calculation

Threshold Neuron: A Brain-inspired Artificial Neuron for Efficient On-device Inference

Spike2Former: Efficient Spiking Transformer for High-performance Image Segmentation

Event-based backpropagation on the neuromorphic platform SpiNNaker2

Numerical analysis and simulation of lateral memristive devices: Schottky, ohmic, and multi-dimensional electrode models