Non-convergence of stochastic gradient descent in the training of deep neural networks.

Patrick Cheridito Arnulf Jentzen Florian Rossmannek

Published in: CoRR (2020)

Keyphrases

stochastic gradient descent
neural network
least squares
loss function
step size
early stopping
matrix factorization
training speed

convergence rate
random forests
training algorithm
training process
convergence speed
support vector machine
weight vector

regularization parameter
back propagation
multiple kernel learning
importance sampling
online algorithms
linear svm
supervised learning

iterative algorithms
support vector
collaborative filtering
training data
learning rate
logistic regression
online learning