On the Inherent Regularization Effects of Noise Injection During Training.
Oussama DhifallahYue M. LuPublished in: CoRR (2021)
Keyphrases
- signal to noise ratio
- early stopping
- data sets
- training and testing data
- noisy data
- noise level
- training process
- random noise
- missing data
- impulse noise removal
- stochastic gradient descent
- noisy environments
- image noise
- parameter selection
- noise removal
- gaussian noise
- training algorithm
- noise reduction
- prior information
- training examples
- training samples
- denoising
- supervised learning
- multiscale