Unifying Adversarial Training Algorithms with Flexible Deep Data Gradient Regularization
This work provides a unifying framework for adversarial training in deep learning, which is incremental as it simplifies and extends prior methods.
The paper introduces DataGrad, a framework that unifies various adversarial training methods for deep neural networks by showing they are instances of a regularized objective, and demonstrates that it outperforms other regularization forms on both original and adversarial datasets, with multi-task integration yielding the most robust performance.
Many previous proposals for adversarial training of deep neural nets have included di- rectly modifying the gradient, training on a mix of original and adversarial examples, using contractive penalties, and approximately optimizing constrained adversarial ob- jective functions. In this paper, we show these proposals are actually all instances of optimizing a general, regularized objective we call DataGrad. Our proposed DataGrad framework, which can be viewed as a deep extension of the layerwise contractive au- toencoder penalty, cleanly simplifies prior work and easily allows extensions such as adversarial training with multi-task cues. In our experiments, we find that the deep gra- dient regularization of DataGrad (which also has L1 and L2 flavors of regularization) outperforms alternative forms of regularization, including classical L1, L2, and multi- task, both on the original dataset as well as on adversarial sets. Furthermore, we find that combining multi-task optimization with DataGrad adversarial training results in the most robust performance.