Neural networks trained with SGD learn distributions of increasing complexity.
Maria RefinettiAlessandro IngrossoSebastian GoldtPublished in: CoRR (2022)
Keyphrases
- neural network
- multilayer perceptron
- fuzzy logic
- auto associative
- genetic algorithm
- trained neural networks
- pre trained
- training set
- artificial neural networks
- worst case
- multi layer perceptron
- pattern recognition
- feed forward
- random variables
- trained neural network
- backpropagation algorithm
- data sets
- back propagation
- computational complexity
- recurrent neural networks
- fuzzy rules
- probability distribution
- machine learning
- number of hidden units
- average case complexity