On regularization of gradient descent, layer imbalance and flat minima.
Boris GinsburgPublished in: CoRR (2020)
Keyphrases
- cost function
- multi layer
- scale space
- objective function
- loss function
- mumford shah functional
- stochastic gradient descent
- class distribution
- decision trees
- least squares
- regularization parameter
- application layer
- middle layer
- neural network
- saddle points
- imbalanced datasets
- piecewise smooth
- blind deconvolution
- reproducing kernel hilbert space
- class imbalance
- total variation
- active contours