HPC

We present ReLATE, a deep reinforcement learning framework for automatically constructing an efficient sparse tensor format

We ask whether AI is capable of handling the complexity of sparse computation on their own. Unfortunately (or fortnuately?) As of today, the answer is an astounding no. However, we can still leverage AI to augment human expertise.

We present a custom geometric multigrid preconditioned conjugate gradient method that applies summation-by-parts(SBP)-preserving interpolations and a custom matrix-free GPU kernel that achieves up to 5x speedup compared to solvers from PETSc and AmgX

We demonstrate the efficiency and performance-portability of encoding sparse tensors in a linearized format

We discuss using high-performance tensor decomposition for topic modeling and malware detection

We discuss our experience in optimizing the CP and Tucker decomposition algorithms for sparse datasets on a distributed system.

We discuss our experience in optimizing the CP and Tucker decomposition algorithms for sparse datasets on a distributed system.

We share our experience in using tensor decomposition for detecting malware

We discuss our experience in optimizing the non-negative Tucker decomposition for sparse datasets on a distributed system.

We discuss our experience in optimizing the CP and Tucker decomposition algorithms for sparse datasets on a distributed system.