[Colloquium] REMINDER: IDEAL Seminar 4/14: Ali Vakilian, University of Wisconsin—Madison

[Alicia McClarin]

*When:*     Tuesday, April 14th at 3:00pm

*Where:*    https://zoom.us/j/726610211?pwd=MWEwTGRkM1pvWHJ0ZEkvQ2xoZVFGQT09


*Who: *      Ali Vakilian, University of Wisconsin—Madison


*Title: *      New Aspects of Algorithms for Massive Data

*Abstract:* The early line of research on processing large datasets mainly
focused on defining new computational models suitable for massive data
analysis (e.g., streaming, sublinear and parallel) and understanding the
complexity of problems in those models (e.g., space, query and round
complexity). However, in recent years and due to the success of machine
learning in applications, new challenges and directions have been
introduced in the design of algorithms for massive data. In this talk, I
will focus on two of these new aspects.

The first aspect is to use “machine-learned advice” to improve the
performance of algorithms which is referred to as “learning-based” (aka
data driven) algorithms. I will consider two basic problems in massive data
analysis, namely frequency estimation and low-rank approximation. For these
problems, we present learning-based algorithms that are augmented with
learned oracles. Our algorithms improve upon the state-of-the-art, both
provably and empirically, if the learned oracle has “high-accuracy”.
Besides, they achieve the same asymptotic accuracy as the best known
“non-learned” algorithms even if the oracle performs poorly.

The second aspect is to incorporate “fairness” in algorithm design. Machine
learning algorithms are now used in many key decision making applications
relying on large volumes of available data. Due to growing concerns about
creating biases toward a specific population or feature by these automated
approaches, the design of fair algorithms becomes more crucial when dealing
with massive data. In this part, I describe how to design scalable and
practical fair algorithms for the basic task of clustering under several
clustering and fairness objectives.

-- 
*Alicia McClarin*
*Toyota Technological Institute at Chicago*
*6045 S. Kenwood Ave., **Office 504*
*Chicago, IL 60637*
*773-834-3321*
*www.ttic.edu* <http://www.ttic.edu/>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.cs.uchicago.edu/pipermail/colloquium/attachments/20200413/216dddf6/attachment.html>