1: \begin{abstract}

2: This paper focuses on recovering a low-rank tensor from its incomplete measurements. We propose a novel algorithm termed the Single Mode Quasi Riemannian Gradient Descent (SM-QRGD). By exploiting the benefits of both fixed-rank matrix tangent space projection in Riemannian gradient descent and sequentially truncated high-order singular value decomposition (ST-HOSVD), SM-QRGD achieves a much faster convergence speed than existing state-of-the-art algorithms. Theoretically, we establish the convergence of SM-QRGD through the Tensor Restricted Isometry Property (TRIP) and the geometry of the fixed-rank matrix manifold. Numerically, extensive experiments are conducted, affirming the accuracy and efficacy of the proposed algorithm.

3: 

4: % This paper addresses the low-rank tensor recovery problem, focusing on reconstructing tensors with low multilinear rank from substantially incomplete measurements. We propose a novel algorithm, single mode quasi Riemannian gradient descent (SM-QRGD), which enhances computational efficiency compared to traditional methods like iterative hard thresholding and Riemannian gradient descent, specifically in thresholding tensors into low multilinear rank sets. This improvement is achieved by combining fixed-rank matrix tangent space projection and sequentially truncated high-order singular value decomposition (ST-HOSVD). Additionally, we offer recovery guarantees for SMQRGD, whether employing constant or normalized step sizes, based on the tensor restricted isometry Property (TRIP) and the geometry of fixed-rank matrix manifolds. Our extensive numerical experiments validate the superior efficacy of our algorithm compared to current methods.

5: 

6: \end{abstract}

7: