HPC formulations of optimization algorithms for tensor completion.
Parallel Computing(2018)
摘要
•Parallel formulations for ALS, SGD, and CCD++ for tensor completion are presented.•CCD++ is shown to scale through 16384 cores.•Randomization accelerates the convergence rates of ALS and CCD++.•The best algorithm depends on both dataset and parallel architecture.
更多查看译文
关键词
Sparse tensor,Tensor completion,Unstructured algorithm,Machine learning,Factorization,Recommender system
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络