Kernel Regression Improves Multi‑Way Data Imputation for Graph Flows

KReTTaH, a kernel‑regression and tensor‑train approach, improves multi‑way imputation for dynamic graph flows, beating two baselines in tests September 2025. Read more: getnews.me/kernel-regression-improv... #kernelregression #tensortrain

Tensor Train Completion from Single‑Mode Fiberwise Observations

Method completes tensors from single‑mode fiberwise observations using only standard linear‑algebra steps, guaranteeing provable recovery without random sampling. getnews.me/tensor-train-completion-... #tensorcompletion #tensortrain #fibermode

Tensor Decomposition Boosts Robustness of Vision‑Language Models

A new study shows a lightweight Tensor Train decomposition layer can recover CLIP performance after adversarial attacks, boosting recall@1 on Flickr30K from 7.5% to 19.8%. getnews.me/tensor-decomposition-boo... #tensortrain #clip #multimodal

Adaptive Tensor-Train Decomposition Improves Network Compression

The new LWIQ method cuts tensor‑train rank‑search time by 63.2% and yields a model 3.2× smaller with only a 0.86% drop in top‑1 accuracy on CIFAR‑10 ResNet‑56. getnews.me/adaptive-tensor-train-de... #modelcompression #tensortrain #deeplearning

Tensor Train Isogeometric Solver Accelerates 3D Poisson Simulations

TT‑IGA, a 3‑D isogeometric solver using tensor‑train decomposition, cuts memory use and floating‑point ops while solving the Poisson equation accurately. getnews.me/tensor-train-isogeometri... #tensortrain #isogeometricanalysis

Low-Rank Tensor-Train Finite Difference Method for 3D Parabolic PDEs

Preprint (12 Sep 2025) presents a low‑rank tensor‑train finite‑difference method for 3D parabolic PDEs, cutting memory use by orders of magnitude with Crank‑Nicolson time stepping. getnews.me/low-rank-tensor-train-fi... #tensortrain #parabolicpde