Distributed-memory tensor completion for generalized loss functions in python using new sparse tensor kernels

Tensor computations are increasingly prevalent numerical techniques in data science, but pose unique challenges for high-performance implementation. We provide novel algorithms and systems infrastructure which enable efficient parallel implementation of tensor completion with generalized loss functions. Specifically, we consider alternating minimization, coordinate a quasi-Newton (generalized Gauss-Newton) method. By extending the Cyclops library, implement all these methods high-level Python syntax. To make possible very sparse tensors, introduce new multi-tensor primitives, specialized implementations. compare routines to pairwise contraction tensors by reduction hypersparse matrix formats, find that more theoretical cost execution time experiments. microbenchmarking results on Stampede2 supercomputer demonstrate efficiency primitives functionality. then study performance synthetic 10 billion nonzeros Netflix dataset, considering both least squares Poisson