Faster & Deterministic FPT Algorithm for Worst-Case Tensor Decomposition
Bibliographic record
Abstract
We present a deterministic 2^{k^{𝒪(1)}} poly(n,d) time algorithm for decomposing d-dimensional, width-n tensors of rank at most k over ℝ and ℂ. This improves upon the previous randomized algorithm of Peleg, Shpilka, and Volk (ITCS '24) that takes 2^{k^{k^{𝒪(k)}}} poly(n,d) time and the deterministic n^k^k time algorithms of Bhargava, Saraf, and Volkovich (STOC '21). Our work resolves an open question asked by Peleg, Shpilka, and Volk (ITCS '24) on whether a deterministic Fixed Parameter Tractable (FPT) algorithm exists for worst-case tensor decomposition. We also make substantial progress on the fundamental problem of how the tractability of tensor decomposition varies as the tensor rank increases. Our result implies that we can achieve deterministic polynomial-time decomposition as long as the rank of the tensor is at most (log n)^{1/C}, where C is some fixed constant independent of n and d. Further, we note that there cannot exist a polynomial-time algorithm for k = ω(log n) unless ETH fails. Our algorithm works for all fields; however, the time complexity worsens to 2^{k^{k^{𝒪(1)}}} and requires randomization for finite fields of large characteristics. Both conditions are provably necessary unless there are improvements in the state of the art for system solving over the corresponding fields. Our approach achieves this by designing a proper learning (reconstruction) algorithm for set-multilinear depth-3 arithmetic circuits. On a technical note, we design a "partial" clustering algorithm for set-multilinear depth-3 arithmetic circuits that lets us isolate a cluster from any set-multilinear depth-3 circuit while preserving the structure of the circuit.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.009 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.002 | 0.003 |
| Bibliometrics | 0.001 | 0.002 |
| Science and technology studies | 0.002 | 0.002 |
| Scholarly communication | 0.004 | 0.007 |
| Open science | 0.004 | 0.003 |
| Research integrity | 0.002 | 0.004 |
| Insufficient payload (model declined to judge) | 0.012 | 0.007 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".