[Club2] Invitation: How robust is current machine learning? A perspective on ... @ Thu Dec 20, 2018 14:00 - 15:00 (CET) (club2 at mailbroy.informatik.tu-muenchen.de)

[julianbrunner at gmail.com]

You have been invited to the following event.

Title: How robust is current machine learning? A perspective on overfitting  
and (adversarial) distribution shifts

About the speaker: Ludwig Schmidt is a postdoctoral researcher at UC  
Berkeley working with Moritz Hardt, Ben Recht, and Martin Wainwright.  
Ludwig’s research interest revolve around the foundations of machine  
learning, with the goal of making machine learning more reliable. Before  
Berkeley, Ludwig completed his PhD at MIT under the supervision of Piotr  
Indyk. Ludwig received a Google PhD Fellowship, a Simons-Berkeley research  
fellowship, a best paper award at the International Conference on Machine  
Learning (ICML), and the Sprowls dissertation award from MIT.

Abstract:
Machine learning is now being deployed in safety- and security-critical  
systems such as autonomous vehicles, medical devices, and large recommender  
systems. If we want to use machine learning in these scenarios responsibly,  
we need to understand how reliable our current methodology is. One  
potential danger in the common ML workflow is the repeated use of the same  
test set for parameter tuning. To investigate this issue, I will present  
results of a reproducibility study on the popular CIFAR-10 dataset.  
Surprisingly, we find no signs of overfitting despite multiple years of  
adaptive classifier tuning. Nevertheless, our results show that current  
classifiers are already susceptible to benign shifts in distribution.

In the second part of the talk, I will then describe how robust  
optimization can address some of the challenges arising from distribution  
shifts in the form of adversarial examples. By exploring the loss landscape  
of min-max problems in deep neural networks, we can train classifiers with  
state-of-the art robustness to l_infinity perturbations and small spatial  
transformations.

Based on joint works with Logan Engstrom, Aleksander Madry, Aleksandar  
Makelov, Benjamin Recht, Rebecca Roelofs, Vaishaal Shankar, Brandon Tran,  
Dimitris Tsipras, and Adrian Vladu.

Corresponding papers: https://arxiv.org/abs/1806.00451,  
https://arxiv.org/abs/1706.06083, https://arxiv.org/abs/1712.02779
When: Thu Dec 20, 2018 14:00 – 15:00 Central European Time - Berlin
Where: MI 00.09.038 (Turing)
Calendar: club2 at mailbroy.informatik.tu-muenchen.de
Who:
     * Julian Brunner - creator
     * club2 at mailbroy.informatik.tu-muenchen.de
     * Ludwig Schmidt

Event details:  
https://www.google.com/calendar/event?action=VIEW&eid=MWFldGU2ZWVuaDRiNGxtdGxuYnVtM2M1cGEgY2x1YjJAbWFpbGJyb3kuaW5mb3JtYXRpay50dS1tdWVuY2hlbi5kZQ&tok=NTIjc2U2ZWJlM3RvZmY0Y2g1bm11bmlibTVtOThAZ3JvdXAuY2FsZW5kYXIuZ29vZ2xlLmNvbTdjNDE0ODY2NmExMTZmODhiMzNlYjJlMGNjOGVjYmUwNWRkNGRiMTE&ctz=Europe%2FBerlin&hl=en&es=0

Invitation from Google Calendar: https://www.google.com/calendar/

You are receiving this courtesy email at the account  
club2 at mailbroy.informatik.tu-muenchen.de because you are an attendee of  
this event.

To stop receiving future updates for this event, decline this event.  
Alternatively you can sign up for a Google account at  
https://www.google.com/calendar/ and control your notification settings for  
your entire calendar.

Forwarding this invitation could allow any recipient to modify your RSVP  
response. Learn more at  
https://support.google.com/calendar/answer/37135#forwarding
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailmanbroy.informatik.tu-muenchen.de/pipermail/club2/attachments/20181210/a8410b38/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: text/calendar
Size: 3769 bytes
Desc: not available
URL: <https://mailmanbroy.informatik.tu-muenchen.de/pipermail/club2/attachments/20181210/a8410b38/attachment-0001.ics>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: invite.ics
Type: application/ics
Size: 3838 bytes
Desc: not available
URL: <https://mailmanbroy.informatik.tu-muenchen.de/pipermail/club2/attachments/20181210/a8410b38/attachment-0001.bin>