- 일정 및 장소: 3/6(목) 16:00, AS301
- 제목: Shaping the Future of Bio-Inspired Intelligence: Energy-Efficient Neural Network Accelerators for Millimeter-Scale Systems
- 연사: 이인희 교수, University of Pittsburgh
- 내용: As bio-inspired intelligence becomes increasingly vital for next-generation systems, miniature sensing and imaging platforms face the challenge of achieving high-performance neural processing within extreme size and power constraints. This talk presents recent advances in neural network accelerator designs tailored for millimeter-scale systems. We will explore four state-of-the-art hardware solutions: a dynamic neural fields accelerator for object tracking, an ultra-low-power CNN layer accelerator, an energy-aware adaptive inference system with multi-exit strategies, and a ternary-input, binary-weight CNN accelerator optimized for object classification. These designs demonstrate substantial improvements in processing speed, energy efficiency, and accuracy while maintaining compact footprints suitable for energy-autonomous miniature devices. We will discover how these innovations contribute to the development of bio-inspired intelligent systems and the future of semiconductor technologies.
- 주관: 홍성완 교수 연구실