[Eecs_phd] ML Tech Talk - Today - Big Data Club

[Hunter, Tiffany]

Topics: Adversarial Attack - Machine Learning
Title: Spies and Spies Alone
Abstract: In recent years, deep learning neural network models are increasingly
being deployed in safety and mission critical systems. However, such
models are known to be susceptible to "adversarial examples" -- data
examples that have been modified with tiny perturbations, often
imperceptible to humans -- which reliably fool a network into, e.g.,
predicting the wrong classification label. In this talk we cover a
brief history of research in this area from the discovery of
adversarial examples in 2013 to the present day. We also present and
analyze the performance of a simple ensemble learning method for
detecting adversarial examples.
Presenter: Alexander Bagnall, CS Graduate Student
Date: Thursday, 3-22-2018
Time: 6 PM - 7:30 PM
Location: Academic & Research Center (ARC) 121
Pizza and water will be provided.
Contact: Yichao Li, president of Big Data Club, yl079811 at ohio.edu<mailto:yl079811 at ohio.edu>.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.ohio.edu/pipermail/eecs_phd/attachments/20180322/b45b1054/attachment.html>