"Oddball SGD": Novelty Driven Stochastic Gradient Descent for Training Deep Neural Networks.
Andrew J. R. SimpsonPublished in: CoRR (2015)
Keyphrases
- stochastic gradient descent
- neural network
- least squares
- loss function
- step size
- training speed
- matrix factorization
- early stopping
- random forests
- support vector machine
- importance sampling
- weight vector
- online algorithms
- support vector
- alternating least squares
- regularization parameter
- training process
- multiple kernel learning
- convergence speed
- genetic algorithm
- training algorithm
- machine learning algorithms
- recommender systems
- feature selection