Finite Depth and Width Corrections to the Neural Tangent Kernel.
Boris HaninMihai NicaPublished in: ICLR (2020)
Keyphrases
- network architecture
- neural network
- feature space
- kernel methods
- depth map
- support vector
- artificial neural networks
- linear separability
- kernel function
- kernel regression
- nonlinear predictive control
- data sets
- convolution kernel
- kernel machines
- kernel pca
- neural model
- bio inspired
- depth information
- learning rules
- positive definite
- associative memory
- biologically plausible
- input space
- similarity function
- reinforcement learning