1: \begin{abstract}

2: We study the best low-rank Tucker decomposition of symmetric tensors, advocating a straightforward projected gradient descent (PGD) method for its computation. 

3: The main application of interest is in decomposing higher-order multivariate moments, which are symmetric tensors.

4: We develop scalable adaptations of the basic PGD method and higher-order eigenvalue decomposition (HOEVD) to decompose sample moment tensors.

5: With the help of implicit and streaming techniques, we evade the overhead cost of building and storing the moment tensor.

6: Such reductions make computing the Tucker decomposition realizable for large data instances in high dimensions. 

7: Numerical experiments demonstrate the efficiency of the algorithms and the  applicability of moment tensor decompositions to real-world datasets.

8: Last, we study the convergence on the Grassmannian manifold, and prove that the update sequence derived by the PGD solver achieves first and second-order criticality. 

9: \end{abstract}

10: