Defend Deep Neural Networks Against Adversarial Examples via Fixed andDynamic Quantized Activation Functions.
Adnan Siraj RakinJinfeng YiBoqing GongDeliang FanPublished in: CoRR (2018)
Keyphrases
- activation function
- neural network
- artificial neural networks
- feed forward neural networks
- back propagation
- feed forward
- neural nets
- hidden layer
- connection weights
- neural architecture
- learning rate
- hidden neurons
- feedforward neural networks
- multilayer perceptron
- network architecture
- radial basis function
- training examples
- fuzzy neural network
- hidden nodes
- basis functions
- fuzzy logic
- pattern recognition
- neural network model
- training phase
- multi layer perceptron
- multi layer
- random variables
- active learning
- support vector
- training data
- genetic algorithm