Less is More: Culling the Training Set to Improve Robustness of Deep Neural Networks
By Yongshuai Liu, Jiyu Chen, Hao Chen. 2018
14 citations recorded by the source corpus.
Deep neural networks are vulnerable to adversarial examples. Prior defenses attempted to make deep networks more robust by either improving the network architecture or adding adversarial examples into the training set, with their respective limitations. We propose a new direction. Motivated by recent research that shows that outliers in the training set have a high negative influence on the trained model, our approach makes the model more robust by detecting and removing outliers in the training set without modifying the network architecture or requiring adversarial examples. We propose two methods for detecting outliers based on canonical examples and on training errors, respectively. After removing the outliers, we train the classifier with the remaining examples to obtain a sanitized model. Our evaluation shows that the sanitized model improves classification accuracy and forces the attacks to generate adversarial examples with higher distortions. Moreover, the Kullback-Leibler divergence from the output of the original model to that of the sanitized model allows us to distinguish between normal and adversarial examples reliably.
Subfields: Artificial Intelligence, Safety, Robustness & Interpretability
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Nicholas Carlini
- Nicolas Papernot
- Florian Tramèr
- Patrick Pérez
- Dawn Song
- Ú. Erlingsson
- Ciarán Hughes
- Karl Amende
In Safety, Robustness & Interpretability
- Nicholas Carlini
- Nicolas Papernot
- Dawn Song
- Patrick Pérez
- Ú. Erlingsson
- Florian Tramèr
- Senthil Kumar Yogamani
- Michal Uřičář
Citations in the corpus
- WoodScape: A Multi-Task, Multi-Camera Fisheye Dataset for Autonomous Driving (2019, background): “However, at the same time, it was shown that careful inspection of the training sets for outliers improves the robustness of deep neural networks [36], especially with regards to the adversarial examples.”
- Distribution Density, Tails, and Outliers in Machine Learning: Metrics and Applications (2019, background): “In fact, prior work has found that discarding outliers from the training data can help with both classifying and detecting adversarial examples (Liu et al., 2018).”