Скачать с ютуб ASPLOS'24 - Session 9C - ML Systems and Optimizations в хорошем качестве

ASPLOS'24 - Session 9C - ML Systems and Optimizations

ASPLOS'24: The International Conference on Architectural Support for Programming Languages and Operating Systems Session 9C: ML Systems and Optimizations Session Chair: Sangeeta Chowdhary (AMD Research) Papers Presented: 1. SmartMem: Layout Transformation Elimination and Adaptation for Efficient DNN Execution on Mobile Authors: Wei Niu, Md Musfiqur Rahman Sanim, and Zhihao Shu (University of Georgia);Jiexiong Guan (William & Mary);Xipeng Shen (North Carolina State University);Miao Yin (University of Texas at Arlington);Gagan Agrawal (University of Georgia);Bin Ren (William & Mary) 2. Dr. DNA: Combating Silent Data Corruptions in Deep Learning using Distribution of Neuron Activations Authors: Dongning Ma (Villanova University);Fred Lin, Alban Desmaison, Joel Coburn, Daniel Moore, and Sriram Sankar (Meta); Xun Jiao (Villanova University and Meta) 3. GPU-based Private Information Retrieval for On-Device Machine Learning Inference Authors: Maximilian Lam (Harvard University); Jeff Johnson (Meta); Wenjie Xiong (Virginia Tech); Kiwan Maeng (Pennsylvania State University); Udit Gupta (Harvard University); Yang Li, Liangzhen Lai, and Ilias Leontiadis (Meta); Minsoo Rhu (KAIST and Meta); Hsien-Hsin S. Lee (Intel); Vijay Janapa Reddi, Gu-Yeon Wei, and David Brooks (Harvard University); Edward Suh (Meta and Cornell University) 4. [Distinguished Artifact Evaluation Award] RECom: A Compiler Approach to Accelerate Recommendation Model Inference with Massive Embedding Columns Authors: Zaifeng Pan (Renmin University of China);Zhen Zheng (Alibaba);Feng Zhang and Ruofan Wu (Renmin University of China);Hao Liang (Alibaba);Dalin Wang (Renmin University of China);Xiafei Qiu, Junjie Bai, and Wei Lin (Alibaba);Xiaoyong Du (Renmin University of China) 5. NDPipe: Exploiting Near-data Processing for Scalable Inference and Continuous Training in Photo Storage Authors: Jungwoo Kim and Seonggyun Oh (DGIST);Jaeha Kung (Korea University);Yeseong Kim and Sungjin Lee (DGIST)