[Colloquium] Yao/MS Presentation/Apr 5, 2019

[Margaret Jaffey]

This is an announcement of Yuanshun (Kevin) Yao's MS Presentation.

------------------------------------------------------------------------------
Date:  Friday, April 5, 2019

Time:  12:00 PM

Place:  Ryerson 276

M.S. Candidate:  Yuanshun Yao

M.S. Paper Title: Empirical Analysis of Machine Learning as a Service

Abstract:
Machine learning (ML) today plays an integral role in a wide range of
critical applications. However, deploying and maintaining an ML system
is notoriously complicated and resource consuming in terms of
management. In the absence of instructions on building custom-tailored
ML systems, ML as a Service (MLaaS) is an attractive alternative.
These services run on the cloud, and provide a query interface to an
ML classifier trained on uploaded datasets.

MLaaS today are opaque systems, with little known about their
efficacy, their underlying mechanisms and relative merits. In this
study, we offer a first look at empirically quantifying performance of
six of the most popular MLaaS platforms across a large number of
labeled datasets for binary classification. We seek to understand how
MLaaS systems compare in performance against each other, and against a
fully customized and tuned local ML library. Additionally, we wish to
better understand the correlations between complexity, performance and
performance variability. Last but not least, we want to understand
which key knobs have the biggest impact on performance, and try to
design generalized techniques to optimize those knobs.

Yuanshun's advisors are Prof. Ben Zhao and Prof. Heather Zheng

Login to the Computer Science Department website for details:
 https://newtraell.cs.uchicago.edu/phd/ms_announcements#ysyao

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Margaret P. Jaffey            margaret at cs.uchicago.edu
Department of Computer Science
Student Support Rep (Ry 156)               (773) 702-6011
The University of Chicago      http://www.cs.uchicago.edu
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=