Developing a Loss Prediction-based Asynchronous Stochastic Gradient Descent Algorithm for Distributed Training of Deep Neural Networks.
Junyu LiLigang HeShenyuan RenRui MaoPublished in: ICPP (2020)
Keyphrases
- stochastic gradient descent
- training process
- early stopping
- training algorithm
- cost function
- optimization algorithm
- probabilistic model
- least squares
- worst case
- monte carlo
- neural network
- back propagation
- training speed
- random forests
- step size
- loss function
- particle swarm optimization
- optimal solution
- objective function
- prediction accuracy
- closed form
- convex hull
- matrix factorization
- em algorithm
- support vector
- importance sampling
- learning algorithm
- machine learning