[AI Seminar] AI Seminar sponsored by Fortive on March 03 (NSH 3305) -- Dravyansh Sharma -- Learning piecewise Lipschitz functions in changing environments

Aayush Bansal aayushb at cs.cmu.edu
Sun Mar 1 08:36:47 EST 2020


Dravyansh Sharma will be giving a seminar on "Learning piecewise Lipschitz
functions in changing environments" from *12:00 - 01:00 PM* on March 03
in Newell Simon Hall (NSH) 3305.

CMU AI Seminar is sponsored by Fortive. Lunch will be served.

Following are the details of the talk:

*Title: *Learning piecewise Lipschitz functions in changing environments

*Abstract: *When using machine learning algorithms, often one needs to tune
hyperparameters optimized for a given data instance. For many problems,
including clustering, a small tweak to the parameters can cause a cascade
of changes in the algorithm's behavior, so the algorithm's performance is a
discontinuous function of the parameters. Optimization in the presence of
sharp (non-Lipschitz), unpredictable (w.r.t. time and amount) changes is a
challenging and largely unexplored problem of great significance.

We consider the class of piecewise Lipschitz functions, which is the most
general online setting considered in the literature for the problem. To
capture changing environments, we look at the 'shifting regret', which
allows for a finite number of environment shifts at unknown times. We
provide a shifting regret bound for well-dispersed functions, where
dispersion roughly quantifies the rate at which discontinuities appear in
the utility functions in expectation. Our near-tight lower bounds further
show how dispersion is necessary and sufficient for low regret. We
empirically demonstrate a key application of our algorithms to online
clustering problems on popular benchmarks.

Joint work with Nina Balcan and Travis Dick.

*Bio*: Dravy <http://www.cs.cmu.edu/~dravyans/> is a graduate student at
CMU advised by Nina Balcan <https://www.cs.cmu.edu/~ninamf/>. He is
interested in designing algorithms for machine learning with strong and
provable performance guarantees. Previously he has worked with the Speech
team at Google and completed his undergraduate studies at IIT Delhi. The
work <https://arxiv.org/pdf/1907.09137.pdf> presented in the present talk
has been accepted for publication at AISTATS 2020 <https://www.aistats.org/>,
Palermo, and was awarded the first prize in poster competition at
YinzOR 2019 <https://cmuinforms.org/posts/2019/08/24/yinzor-2019.html>,
Pittsburgh.

To learn more about the seminar series, please visit the website
<http://www.cs.cmu.edu/~aiseminar/>.

-- 
Aayush Bansal
http://www.cs.cmu.edu/~aayushb/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20200301/0d84a91b/attachment.html>


More information about the ai-seminar-announce mailing list