PROGRAM - International Workshop on Performance Analysis of Machine Learning Systems

Sunday, August 23, 2020 – Virtual (9:00 am - 5:00 pm US EDT = 13:00 - 21:00 UTC)

FastPath’2020 – In conjunction with ISPASS 2020

Youtube Recording of Entire Workshop (Recording offset times of individual talks are below)

Program

Time Recording Offset Speaker Affiliation Talk Title
9:00 - 9:15 0:00:00 Erik Altman, Parijat Dube, Vijay Janapa Reddi FastPath Chairs Welcome
9:15 - 10:00 0:07:55 Vivienne Sze (*) MIT How to Evaluate Efficient Deep Neural Network Approaches
10:00 - 10:45 0:57:50 Grigori Fursin (*) cKnowledge Enabling Reproducible ML&Systems Research: The Good, the Bad and the Ugly
10:45 - 11:00     Break  
11:00 - 11:30 1:48:40 Bochen Guan University of Wisconsin SpecNet: Spectral Domain Convolutional Neural Network
11:30 - 12:15 2:17:05 Yuhao Zhu (*) University of Rochester Getting Computer Systems Ready for Visual Computing in Ten Years
12:15 - 1:00     Lunch  
1:00 - 1:45 3:08:35 Anju Kambadur, Ania Musial, Ian Hummel (*) Bloomberg L.P. The Bloomberg Data Science Platform
1:45 - 2:30 3:50:48 Chuang Gan (*) MIT-IBM Watson AI Lab Multimodal Intelligence
2:30 - 2:45     Break  
2:45 - 3:15 4:42:10 Ted Pyne Harvard University Quantifying the impact of data encoding on DNN fault tolerance
3:15 - 4:00 5:04:15 Colby Banbury (*) Harvard University tinyMLPerf: Benchmarking Ultra-low Power Machine Learning Systems
4:00 - 4:15     Break  
4:15 - 5:00 5:39:00 Leopold Grinberg (*) IBM Research Unsupervised Training and Unified Global Address Spaces

* = Invited

Technical Papers

   
Title SpecNet: Spectral Domain Convolutional Neural Network
Authors Bochen Guan, Jinnian Zhang, William A. Sethares, Richard Kijowski (University of Wisconsin), Fang Liu (Harvard University)
   
Title Quantifying the impact of data encoding on DNN fault tolerance
Authors Edward Pyne, Lillian Pentecoste, Udit Gupta, Gu-Yeon Wei, David Brooks (Harvard University)