I will start with a brief intro to tensors and CP decompositions then get into symmetric tensors, their correspondence to higher-order moments, and the decomposition of symmetric tensors. We developed a method to decompose a symmetric tensor that avoids forming the tensor explicitly, which quickly becomes unreasonable as the size of the problem grows. Our implicit method is more efficient and allows us to compute decompositions to tensors that are too large to form explicitly. I will show examples of using symmetric tensor decompositions to estimate symmetric tensor rank and recover means of Gaussian mixture models. This is joint work with Dr. Tammy Kolda at Sandia National Labs.

Please use this link to attend the virtual seminar:

Bluejeans Link: https://bluejeans.com/807515524/1551