Non-convergence to global minimizers for Adam and stochastic gradient descent optimization and constructions of local minimizers in the training of artificial neural networks.
Arnulf JentzenAdrian RiekertPublished in: CoRR (2024)
Keyphrases
- stochastic gradient descent
- artificial neural networks
- global minimizers
- least squares
- step size
- matrix factorization
- loss function
- convergence rate
- random forests
- global minimum
- online algorithms
- weight vector
- regularization parameter
- energy functional
- cost function
- piecewise constant
- convergence speed
- active contour model
- neural network
- support vector machine
- active contours
- linear svm
- collaborative filtering
- variational framework
- decision trees
- optimal solution
- multi objective
- particle swarm optimization
- online learning
- pairwise
- level lines
- computer vision