Timezone: »
We investigate the robustness of stochastic approximation approaches against data poisoning attacks. We focus on two-layer neural networks with ReLU activation and show that under a specific notion of separability in the RKHS induced by the infinite-width network, training (finite-width) networks with stochastic gradient descent is robust against data poisoning attacks. Interestingly, we find that in addition to a lower bound on the width of the network, which is standard in the literature, we also require a distribution-dependent upper bound on the width for robust generalization. We provide extensive empirical evaluations that support and validate our theoretical results.
Author Information
Yunjuan Wang (Johns Hopkins University)
Poorya Mianjy (Johns Hopkins University)
Raman Arora (Johns Hopkins University)

Raman Arora received his M.S. and Ph.D. degrees in Electrical and Computer Engineering from the University of Wisconsin-Madison in 2005 and 2009, respectively. From 2009-2011, he was a Postdoctoral Research Associate at the University of Washington in Seattle and a Visiting Researcher at Microsoft Research Redmond. Since 2011, he has been with Toyota Technological Institute at Chicago (TTIC). His research interests include machine learning, speech recognition and statistical signal processing.
Related Events (a corresponding poster, oral, or spotlight)
-
2021 Poster: Robust Learning for Data Poisoning Attacks »
Fri. Jul 23rd 04:00 -- 06:00 AM Room Virtual
More from the Same Authors
-
2023 Poster: Faster Rates of Convergence to Stationary Points in Differentially Private Optimization »
Raman Arora · Raef Bassily · Tomás González · Cristobal Guzman · Michael Menart · Enayat Ullah -
2023 Poster: From Adaptive Query Release to Machine Unlearning »
Enayat Ullah · Raman Arora -
2021 Poster: Dropout: Explicit Forms and Capacity Control »
Raman Arora · Peter Bartlett · Poorya Mianjy · Nati Srebro -
2021 Spotlight: Dropout: Explicit Forms and Capacity Control »
Raman Arora · Peter Bartlett · Poorya Mianjy · Nati Srebro -
2020 Poster: FetchSGD: Communication-Efficient Federated Learning with Sketching »
Daniel Rothchild · Ashwinee Panda · Enayat Ullah · Nikita Ivkin · Ion Stoica · Vladimir Braverman · Joseph E Gonzalez · Raman Arora -
2019 Poster: On Dropout and Nuclear Norm Regularization »
Poorya Mianjy · Raman Arora -
2019 Oral: On Dropout and Nuclear Norm Regularization »
Poorya Mianjy · Raman Arora -
2018 Poster: On the Implicit Bias of Dropout »
Poorya Mianjy · Raman Arora · Rene Vidal -
2018 Oral: On the Implicit Bias of Dropout »
Poorya Mianjy · Raman Arora · Rene Vidal -
2018 Poster: Streaming Principal Component Analysis in Noisy Setting »
Teodor Vanislavov Marinov · Poorya Mianjy · Raman Arora -
2018 Poster: Stochastic PCA with $\ell_2$ and $\ell_1$ Regularization »
Poorya Mianjy · Raman Arora -
2018 Oral: Streaming Principal Component Analysis in Noisy Setting »
Teodor Vanislavov Marinov · Poorya Mianjy · Raman Arora -
2018 Oral: Stochastic PCA with $\ell_2$ and $\ell_1$ Regularization »
Poorya Mianjy · Raman Arora