Improving the Adversarial Robustness and Interpretability of Deep Neural Networks by Regularizing Their Input Gradients.
Andrew Slavin RossFinale Doshi-VelezPublished in: AAAI (2018)
Keyphrases
- neural network
- desired output
- multi agent
- pattern recognition
- artificial neural networks
- hidden units
- recurrent neural networks
- fuzzy logic
- prediction accuracy
- real time
- feed forward
- input data
- back propagation
- artificial intelligence
- neural nets
- multi layer
- fuzzy systems
- cellular neural networks
- genetic algorithm
- image segmentation
- rough sets
- data mining
- decision trees
- multilayer perceptron
- image gradient
- deep learning
- information retrieval