Lower error bounds for the stochastic gradient descent optimization algorithm: Sharp convergence rates for slowly and fast decaying learning rates.
Arnulf JentzenPhilippe von WurstembergerPublished in: J. Complex. (2020)
Keyphrases
- optimization algorithm
- error bounds
- learning rate
- convergence rate
- stochastic gradient descent
- step size
- weight vector
- convergence speed
- multi objective
- optimization method
- differential evolution
- worst case
- theoretical analysis
- particle swarm optimization pso
- gaussian kernels
- matrix factorization
- loss function
- least squares
- online algorithms
- linear combination
- special case
- feature space
- random forests
- reinforcement learning
- feature selection
- neural network