Research map / Learning theory and algorithms / Matrix and tensor methods
Computational and Statistical Guarantees for Tensor-on-Tensor Regression With Tensor Train Decomposition
Zhen Qin, Zhihui Zhu
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use TT structure rank (20,20,20) instead of full rank under noise variance 5 (ablation) | IHT · LFW, m=5000, noise 5 | relative test error: 0.523 -> 0.3892 | Improved |
| use TT ranks (20,20,20) instead of full rank under noise variance 5 (ablation) | RGD · LFW, m=5000, noise 5 | relative test error: 0.5509 -> 0.4181 | Improved |
| use TT ranks (30,30,30) instead of full rank under noise variance 1 (ablation) | IHT · LFW, m=5000, noise 1 | relative test error: 0.3422 -> 0.3199 | Improved |
| use TT rank (50,50,50) instead of full rank with noiseless data (ablation) | IHT · LFW, m=5000, noise 0 | relative test error: 0.2957 -> 0.2956 | No clear effect |
| use TT rank (10,10,10) instead of (50,50,50) under noiseless data (ablation) | IHT · LFW, m=5000, noise 0 | relative test error: 0.2956 -> 0.3673 | Hurt |
| use IHT instead of RGD | TT-ToT regression · LFW, m=5000, noise 1, rank (50,50,50) | relative test error: 0.3713 -> 0.3275 | Improved |
| use IHT instead of RGD | TT-ToT regression · LFW, m=10000, noise 0, rank (50,50,50) | relative test error: 0.315 -> 0.2821 | Improved |
| increase training observations from 5000 to 10000 | IHT full rank · LFW, noise 1 | relative test error: 0.3422 -> 0.3056 | Improved |
| increase training observations from 5000 to 10000 | RGD TT ranks (30,30,30) · LFW, noise 5 | relative test error: 0.4365 -> 0.381 | Improved |
| use RGD with TT ranks (50,50,50) instead of IHT regarding per-iteration time | RGD vs IHT · LFW, m=5000 | relative test error: 0.99 -> 0.88 | Mixed |
Closest papers by meaning
- High-Dimensional Tensor Regression With Oracle Properties (ICML 2025 · similarity 0.89)
- Efficient Leverage Score Sampling for Tensor Train Decomposition (NeurIPS 2024 · similarity 0.89)
- Fast Tensor Completion via Approximate Richardson Iteration (ICML 2025 · similarity 0.89)
- Low-Rank Tensor Transitions (LoRT) for Transferable Tensor Regression (ICML 2025 · similarity 0.88)
- Generalized Tensor Decomposition for Understanding Multi-Output Regression under Combinatorial Shifts (NeurIPS 2024 · similarity 0.88)
- Factor Augmented Tensor-on-Tensor Neural Networks (AAAI 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
