Colored Noise Injection for Training Adversarially Robust Neural Networks.
Evgenii ZheltonozhskiiChaim BaskinYaniv NemcovskyBrian ChmielAvi MendelsonAlex M. BronsteinPublished in: CoRR (2020)
Keyphrases
- neural network
- image noise
- training process
- training algorithm
- greater robustness
- pattern recognition
- artificial neural networks
- geometric distortions
- back propagation
- feedforward neural networks
- salt pepper
- noisy environments
- feed forward neural networks
- noise model
- computationally efficient
- training set
- noise level
- multilayer perceptron
- noisy data
- robust statistical
- training phase
- random noise
- imaging artifacts
- online learning
- multilayer neural network
- neural network structure
- neural network training
- recurrent networks
- multi layer
- registration errors
- test set
- noise reduction
- recurrent neural networks