An Adaptive Interval-Constrained Tucker Decomposition for Efficient Multilevel Fast Multipole Algorithm
Abstract
To enhance far-field tensor storage efficiency and computational performance in the multilevel fast multipole algorithm (MLFMA), we propose (i) an adaptive interval-constrained Tucker decomposition that reformulates rank selection under an error interval ϵlower[ϵlower, ϵlower] to avoid overly conservative truncation and eliminate the manual tolerance adjustments, and (ii) a slice precomputation strategy that converts runtime Tucker contractions into matrix–vector products without repeated reshaping overhead to accelerate Tucker-compressed tensor–vector multiplication. Across five PEC-sphere cases, the method achieves 8.0–11.5:1 compression under 0.80-1.00 × 10-3tensor-level relative error and delivers 2.3–6.3× speedup in aggregation and 4.9–7.9× speedup in disaggregation.
Fields of Study & Subjects
Direction-of-Arrival Estimation Techniques
Electromagnetic Scattering and Analysis
Electromagnetic Simulation and Numerical Methods
Publication details
- Journal / source
- IEEE Antennas and Wireless Propagation Letters
- Publisher
- Institute of Electrical and Electronics Engineers (IEEE)
- Volume
- 25
- Issue
- 6
- Pages
- 2561–2565
- Publication date
- 22 April 2026