UNIQ: Uniform Noise Injection for Non-Uniform Quantization of Neural Networks.
Chaim BaskinNatan LissEli SchwartzEvgenii ZheltonozhskiiRaja GiryesAlex M. BronsteinAvi MendelsonPublished in: ACM Trans. Comput. Syst. (2021)
Keyphrases
- neural network
- quantization noise
- uniform quantization
- artificial neural networks
- random noise
- noise level
- pattern recognition
- back propagation
- self organizing maps
- neural network model
- image noise
- genetic algorithm
- noise model
- signal to noise ratio
- neural nets
- image structure
- fuzzy systems
- training process
- additive noise
- arbitrary shape
- multilayer perceptron
- feed forward
- noise free
- color quantization
- noisy data
- code length
- fuzzy logic