Adversarial Examples Are Not Bugs, They Are Features

Channel:
Switzerland Yannic Kilcher
Subscribers:
285,000
Published on ● Video Link: https://www.youtube.com/watch?v=hMO6rbMAPew


Duration: 40:21
10,432 views
358

Abstract:
Adversarial examples have attracted significant attention in machine learning, but the reasons for their existence and pervasiveness remain unclear. We demonstrate that adversarial examples can be directly attributed to the presence of non-robust features: features derived from patterns in the data distribution that are highly predictive, yet brittle and incomprehensible to humans. After capturing these features within a theoretical framework, we establish their widespread existence in standard datasets. Finally, we present a simple setting where we can rigorously tie the phenomena we observe in practice to a misalignment between the (human-specified) notion of robustness and the inherent geometry of the data.

Authors: Andrew Ilyas, Shibani Santurkar, Dimitris Tsipras, Logan Engstrom, Brandon Tran, Aleksander Madry

https://arxiv.org/abs/1905.02175



Other Videos By Yannic Kilcher


2019-08-13Gauge Equivariant Convolutional Networks and the Icosahedral CNN
2019-08-12Processing Megapixel Images with Deep Attention-Sampling Models
2019-08-09Manifold Mixup: Better Representations by Interpolating Hidden States
2019-08-08Learning World Graphs to Accelerate Hierarchical Reinforcement Learning
2019-08-05Reconciling modern machine learning and the bias-variance trade-off
2019-07-05Conversation about Population-Based Methods (Re-upload)
2019-07-03XLNet: Generalized Autoregressive Pretraining for Language Understanding
2019-06-13Talking to companies at ICML19
2019-06-12Population-Based Search and Open-Ended Algorithms
2019-06-10I'm at ICML19 :)
2019-05-14Adversarial Examples Are Not Bugs, They Are Features
2019-05-10Reinforcement Learning, Fast and Slow
2019-05-09S.H.E. - Search. Human. Equalizer.
2019-05-06Blockwise Parallel Decoding for Deep Autoregressive Models
2019-04-27Discriminating Systems - Gender, Race, and Power in AI
2019-02-19The Odds are Odd: A Statistical Test for Detecting Adversarial Examples
2019-02-18Neural Ordinary Differential Equations
2019-02-18GPT-2: Language Models are Unsupervised Multitask Learners
2019-02-02Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
2019-01-30BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
2019-01-09What’s in a name? The need to nip NIPS


Tags:
machine learning
deep learning
adversarial examples
adversarial samples
pgd
projected gradient descent
vulnerabiliby
security
artificial intelligence
MIT
geometry
classifier
deep neural network
attack
convolutional neural networks
research
robust features
robust classifier
robust network
neural network