Efficient Deep Neural Network Inference with Near-Edge Accelerator Offload Hao Liu, Ph.D. Student, Computer Science Nov 11, 09:30 - 11:30 B2 R5220 edge computing Efficient AI accelerators multi-stage optimization algorithm distributed computing Distributed algorithms This thesis develops efficient distributed inference strategies for deep neural networks that exploit near-edge accelerators through edge, near-edge, and cloud tier offloading, optimized model partitioning, and collaborative expert-based processing to achieve low-latency, energy-efficient, and accurate AI inference on resource-constrained edge systems.
Efficient Machine Learning across Hardware, Software, and Algorithms Prof. Mohamed Abdelfattah, Electrical and Computer Engineering Department at Cornell University Dec 17, 14:00 - 15:30 B2 L5 R5209 Efficient machine learning Hardware software algorithms Deep neural networks (DNNs) are revolutionizing computing, necessitating an integrated approach across the computing stack to optimize efficiency. In this talk, I will explore the frontier of DNN optimization, spanning algorithms, software, and hardware. We'll start with hardware-aware neural architecture search, demonstrating how tailoring DNN architectures to specific hardware can drastically enhance performance.
