Improving the Robustness of Quantized Deep Neural Networks to White-Box Attacks using Stochastic Quantization and Information-Theoretic Ensemble Training.
Saurabh FarkyaAswin RaghavanAvi ZiskindPublished in: CoRR (2023)
Keyphrases
- feature selection
- information theoretic
- mutual information
- neural network
- white box
- training process
- information theory
- black box
- competitive learning
- jensen shannon divergence
- theoretic framework
- training set
- watermark embedding
- entropy measure
- digital image watermarking scheme
- information theoretic measures
- information bottleneck
- kullback leibler divergence
- log likelihood
- support vector
- relative entropy
- pattern recognition
- feature space
- minimum description length
- kl divergence
- training data
- watermarking scheme
- machine learning
- back propagation
- ensemble methods
- supervised learning
- test set
- knn
- source code
- bregman divergences
- training samples
- data sets
- test data