On the Interpretability of Regularisation for Neural Networks Through Model Gradient Similarity.
Vincent SzolnokyViktor AnderssonBalázs KulcsárRebecka JörnstenPublished in: CoRR (2022)
Keyphrases
- neural network
- neural network model
- high level
- objective function
- probabilistic model
- network model
- computational model
- pattern recognition
- prior knowledge
- multi layer
- experimental data
- gradient method
- recurrent neural networks
- mathematical model
- parameter estimation
- distance measure
- edge detection
- probability distribution
- similarity measure