Training Quantised Neural Networks with STE Variants: the Additive Noise Annealing Algorithm.
Matteo SpallanzaniGian Paolo LeonardiLuca BeniniPublished in: CoRR (2022)
Keyphrases
- additive noise
- neural network
- annealing algorithm
- spatial domain
- noisy images
- maximum likelihood
- pattern recognition
- random variables
- deterministic annealing
- genetic algorithm
- speech signal
- hidden markov models
- active learning
- image denoising
- global optimization
- training set
- object recognition
- protein folding
- information retrieval