New R-ItCUR Algorithm Enhances Tensor Completion Robustness

Researchers have developed a new algorithm called Robust Iterative t-CUR (R-ItCUR) for completing low-tubal-rank tensors from partial data that may contain significant outliers. This method partitions the sampled tensor cross, applies adaptive Welsch correction for outlier suppression, and updates the low-rank component via projected blockwise gradient descent. R-ItCUR operates directly on the sampled cross, reducing memory and computational demands. Experiments on synthetic data, cardiac MRI, and seismic data show accurate recovery and robustness to gross corruptions, emphasizing the benefit of exploiting cross-concentrated sampling structures. AI

IMPACT Introduces a more robust method for tensor completion, potentially improving data analysis in fields like medical imaging and seismic data processing.

RANK_REASON Research paper detailing a new algorithm for tensor completion. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv stat.ML →

• arXiv
• cardiac magnetic resonance imaging
• R-ItCUR
• Welsch correction

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →