Research map / Learning theory and algorithms / Matrix and tensor methods
Tensor Train Decomposition on TensorFlow (T3F)
Alexander Novikov, Pavel Izmailov, Valentin Khrulkov, Michael Figurnov, Ivan Oseledets
Read the paperCheck your idea against 137,217 papers
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 T3F instead of TTPY (1 object CPU) | TT matmul · matmul benchmark | time (ms): 86.191 -> 9.849 | Improved |
| use T3F instead of TTPY (1 object CPU) | TT matvec · matvec benchmark | time (ms): 11.142 -> 1.190 | Improved |
| use T3F instead of TTPY (1 object CPU) | TT project · project benchmark | time (ms): 116.868 -> 3.001 | Improved |
| use T3F instead of TTPY (1 object CPU) | TT gram · gram benchmark | time (ms): 0.145 -> 0.606 | Hurt |
| move T3F from CPU to GPU (1 object) | T3F matmul · matmul benchmark | time (ms): 9.849 -> 0.950 | Improved |
| move T3F from CPU to GPU (1 object) | T3F project · project benchmark | time (ms): 3.001 -> 13.239 | Hurt |
| move T3F from CPU to GPU (1 object) | T3F gram · gram benchmark | time (ms): 0.606 -> 0.973 | Hurt |
| batch 100 objects on GPU instead of one object on GPU | T3F norm · norm benchmark | time (ms): 1.019 -> 0.044 | Improved |
| batch 100 objects on GPU instead of one object on GPU | T3F gram · gram benchmark | time (ms): 0.973 -> 0.001 | Improved |
| batch 100 objects on CPU instead of one object on CPU | T3F matvec · matvec benchmark | time (ms): 1.190 -> 1.885 | Hurt |
Closest papers by meaning
- On Tensor Train Rank Minimization : Statistical Efficiency and Scalable Algorithm (NeurIPS 2017 · similarity 0.87)
- Einsum Benchmark: Enabling the Development of Next-Generation Tensor Execution Engines (NeurIPS 2024 · similarity 0.84)
- Efficient Leverage Score Sampling for Tensor Train Decomposition (NeurIPS 2024 · similarity 0.83)
- Geometry-aware training of factorized layers in tensor Tucker format (NeurIPS 2024 · similarity 0.83)
- Fast Tensor Completion via Approximate Richardson Iteration (ICML 2025 · similarity 0.82)
- Factor Augmented Tensor-on-Tensor Neural Networks (AAAI 2025 · similarity 0.82)
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.
