[Seminar] Towards Scalable and Efficient Machine Learning as a Service (MLaaS)
Monday, March 21, 2022
11:00 am - 12:00 pm
Assistant Professor of Computer Science and Engineering
University of Nevada, Reno
*Microsoft 365 @cougarnet.uh.edu authentication required to join via Zoom
Driven by the explosive growth of big data, the sustained advances of Machine Learning (ML), and the fast evolving of computer system techniques, the past few years have witnessed a surging demand for Machine Learning as a Service (MLaaS). MLaaS is an emerging computing paradigm that facilitates ML model design, training, inference serving and provides optimized executions of ML tasks in an automated, scalable, and efficient manner.
In this talk, I will demonstrate how to integrate ML algorithm research and system research in synergy to address the pressing challenges in MLaaS. I will first share a story about how our system experience led to a novel large batching algorithm design that revolutionizes large-scale training. Then I will tell another story about how our gradient compression algorithm research helped us to discover overlooked critical features of modern ML systems and thereby build a compression-aware distributed ML system. I will also briefly discuss a promising future of harnessing serverless computing for MLaaS model inference serving. I will conclude my talk with a discussion of interdisciplinary research and future plans.
About the Speaker
Dr. Feng Yan is an Assistant Professor of Computer Science and Engineering at University of Nevada, Reno (UNR) and director of the Intelligent Data and Systems Lab (IDS Lab). Dr. Yan received M.S. and Ph.D. degrees in Computer Science from the College of William and Mary and worked at Microsoft Research and HP Labs. Dr. Yan’s research bridges the fields of big data, machine learning, and systems. The focus of his research is on developing methodologies and building systems that are automated, high-performing, efficient, robust, and user-centric. Some of his recent research topics include large-scale distributed deep learning, machine learning as a service (MLaaS), federated learning, AutoML, serverless computing, and broad topics in cloud and HPC. Dr. Yan is also dedicated to interdisciplinary research and has established fruitful collaborations with domain experts in areas such as health, physics, geography, material science, mechanical engineering, civil engineering, and innovated big data and AI-driven approaches for these domains. Dr. Yan and his team are actively publishing at the most prestigious venues in computer system area (such as SOSP, SC, HPDC, USENIX ATC, EuroSys, FAST, VLDB, etc.) and machine learning area (such as NIPS/NeurIPS, KDD, AAAI, etc.). Dr. Yan and his students are the recipients of the Best Student Paper Award of IEEE CLOUD 2018, the Best Paper Award of CLOUD 2019, and the Best Student Paper Award of ITNG 2021. Dr. Yan is the recipient of the NSF CAREER Award, the NSF CRII Award, the Outstanding Service Award of IEEE ACSOS, the Regents’ Rising Researcher Award, and the CSE Best Researcher Award. Dr. Yan serves as Social Media Chair of ACM SIGMETRICS.