Double-Win Quant: Aggressively Winning Robustness of Quantized Deep Neural Networks via Random Precision Training and Inference.
Yonggan FuQixuan YuMeng LiVikas ChandraYingyan LinPublished in: ICML (2021)
Keyphrases
- neural network
- training process
- training algorithm
- feedforward neural networks
- feed forward neural networks
- structured prediction
- multi layer perceptron
- training set
- back propagation
- artificial neural networks
- genetic algorithm
- neural network training
- backpropagation algorithm
- neural network model
- recurrent networks
- precision and recall
- numerical stability
- neural network structure
- deep architectures
- pattern recognition
- multilayer neural network
- training patterns
- bayesian inference
- error back propagation
- training phase
- feed forward
- high precision
- online learning
- fuzzy systems
- probabilistic inference
- multilayer perceptron
- deep learning
- inference process
- radial basis function
- fault diagnosis
- test set
- fuzzy logic
- feature space
- bayesian networks
- machine learning