A Theory of Non-Linear Feature Learning with One Gradient Step in Two-Layer Neural Networks.
Behrad MoniriDonghwan LeeHamed HassaniEdgar DobribanPublished in: CoRR (2023)
Keyphrases
- neural network
- learning algorithm
- error back propagation
- neural nets
- active learning
- online learning
- learning systems
- multi layer
- knowledge acquisition
- fuzzy logic
- theoretical framework
- learning process
- feature space
- preprocessing step
- feedforward neural networks
- cognitive development
- neural learning
- pattern recognition
- cascade correlation