Quantitative Gaussian Approximation of Randomly Initialized Deep Neural Networks.
Andrea BasteriDario TrevisanPublished in: CoRR (2022)
Keyphrases
- neural network
- gaussian convolution
- linear computational complexity
- series expansion
- fuzzy logic
- pattern recognition
- artificial neural networks
- back propagation
- approximation error
- maximum likelihood
- qualitative and quantitative
- neural network model
- error bounds
- polynomial approximation
- genetic algorithm
- training process
- taylor series expansion
- multi layer
- neural nets
- closed form
- image processing
- density function
- multilayer perceptron
- gaussian mixture
- randomly chosen
- quantitative measures
- self organizing maps
- fuzzy sets
- rough sets
- reinforcement learning