Event-driven Processing and Learning with Spiking Neural Networks
Guardado en:
| Publicado en: | ProQuest Dissertations and Theses (2024) |
|---|---|
| Autor principal: | |
| Publicado: |
ProQuest Dissertations & Theses
|
| Materias: | |
| Acceso en línea: | Citation/Abstract Full Text - PDF |
| Etiquetas: |
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| Resumen: | The drive to engineer technology replicating human capabilities has recently led to the development of event-driven sensors, which emulate the energy-efficient neural mechanisms of biological systems such as the retina and cochlea. Like biological sensors that generate spikes to the changing environment, these bio-inspired event-driven sensors build circuits that dynamically produce the binary events to the changing environmental stimuli. In general, such event-based sensors can achieve higher energy efficiency, better scalability, and lower latency. However, due to the high sparsity and complexity of event-driven data, processing and learning with these sensors remain in their infancy. In this dissertation, we propose to utilize Spiking Neural Networks (SNNs) to tackle event-driven processing and learning. Unlike traditional Artificial Neural Networks (ANNs), SNNs draw inspiration from the brain, processing information in a binary spiking fashion that mirrors natural neural activity. Given the common spiking mechanism between event-driven data and SNNs, it naturally follows that SNNs are well-suited for processing and learning from event-driven data. In this thesis, our exploration focuses on three pivotal areas: 1) Developing SNNs for event-driven classification tasks, including tactile object recognition and slip detection, showcasing their superior performance, energy efficiency, and broad impact. 2) Advancing SNNs for complex event-driven regression challenges like surface normal estimation, demonstrating comparable state-of-the-art accuracy with less energy consumption. 3) Innovating spiking neural architectures by integrating insights from neuroscience, resulting in two models that excel in object recognition with additional robustness and interpretability, respectively. By pushing the boundaries of SNN capabilities and exploring their applications in event-driven processing and learning, this work not only highlights the potential of bio-inspired technologies but also sets the stage for future research in the field. |
|---|---|
| ISBN: | 9798382762890 |
| Fuente: | ProQuest Dissertations & Theses Global |