Fast truncation of mode ranks for bilinear tensor operations
arXiv:1004.491910 citationsh-index: 32
Originality Incremental advance
AI Analysis
It addresses the computational bottleneck of tensor operations for large-scale data analysis, offering a speedup over existing methods.
The paper proposes a fast algorithm for mode rank truncation of bilinear tensor operations, achieving O(nr^3 + r^4) complexity with accuracy limited by square root of machine precision.
We propose a fast algorithm for mode rank truncation of the result of a bilinear operation on 3-tensors given in the Tucker or canonical form. If the arguments and the result have mode sizes n and mode ranks r, the computation costs $O(nr^3 + r^4)$. The algorithm is based on the cross approximation of Gram matrices, and the accuracy of the resulted Tucker approximation is limited by square root of machine precision.