The Regularization Effects of Anisotropic Noise in Stochastic Gradient Descent.
Zhanxing ZhuJingfeng WuBing YuLei WuJinwen MaPublished in: CoRR (2018)
Keyphrases
- stochastic gradient descent
- least squares
- loss function
- step size
- matrix factorization
- regularization parameter
- early stopping
- random forests
- support vector machine
- missing data
- weight vector
- noise level
- importance sampling
- multiple kernel learning
- convergence rate
- cost function
- training set
- convergence speed
- image restoration
- collaborative filtering
- training data
- cross validation
- maximum likelihood
- objective function