Architecture and Learning Optimization of Efficiency-oriented Brain-inspired Large Models

The development of large AI models has driven breakthroughs but raises challenges in efficiency and scalability. This project aims to develop brain-inspired large models (BLMs) to alleviate dependence on computational power and GPUs. Key objectives include building a fully spike-driven BLM for low-power computation, designing efficient learning algorithms for faster training, and implementing BLMs on low-power hardware to evaluate energy efficiency. Additionally, the project explores low-power solutions for brain-inspired computer vision by integrating BLMs with event cameras.