GPUSGD: A GPU-accelerated stochastic gradient descent algorithm for matrix factorization.
Jing JinSiyan LaiSu HuJing LinXiaola LinPublished in: Concurr. Comput. Pract. Exp. (2016)
Keyphrases
- stochastic gradient descent
- matrix factorization
- gpu accelerated
- step size
- loss function
- dynamic programming
- least squares
- learning algorithm
- negative matrix factorization
- collaborative filtering
- convergence rate
- recommender systems
- random forests
- nonnegative matrix factorization
- importance sampling
- k means
- monte carlo
- reinforcement learning
- online algorithms
- similarity measure
- machine learning