VS-Quant: Per-vector Scaled Quantization for Accurate Low-Precision Neural Network Inference.
Steve DaiRangharajan VenkatesanMark RenBrian ZimmerWilliam J. DallyBrucek KhailanyPublished in: MLSys (2021)
Keyphrases
- neural network
- artificial neural networks
- back propagation
- neural network model
- gradient vector
- computationally efficient
- high accuracy
- feature space
- computational complexity
- pattern recognition
- vector space
- belief networks
- successive approximation
- expert systems
- neural network is trained
- high quality
- vector data
- inference process
- quantization error
- feed forward neural networks
- network architecture
- training process
- multi layer
- probabilistic inference
- incremental learning
- feed forward
- data sets
- bayesian networks
- feature vectors