Conservative SPDEs as fluctuating mean field limits of stochastic gradient descent.
Benjamin GessRishabh S. GvalaniVitalii KonarovskyiPublished in: CoRR (2022)
Keyphrases
- stochastic gradient descent
- least squares
- step size
- loss function
- matrix factorization
- markov random field
- random forests
- support vector machine
- regularization parameter
- em algorithm
- bayesian inference
- multiple kernel learning
- online algorithms
- closed form
- belief networks
- weight vector
- importance sampling
- cost function
- parameter estimation
- monte carlo
- lower bound