Sublinear Time Orthogonal Tensor Decomposition
Authors: Zhao Song, David Woodruff, Huan Zhang
NeurIPS 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our implementation shares the same code base 1 as the sketching-based robust tensor power method proposed in [23]. We ran our experiments on an i7-5820k CPU with 64 GB of memory in singlethreaded mode. We ran two versions of our algorithm: the version with pre-scanning scans the full tensor to accurately measure per-slice Frobenius norms and make samples for each slice in proportion to its Frobenius norm in APPROXTIVW; the version without pre-scanning assumes that the Frobenius norm of each slice is bounded by 1 nα T 2 F , α (0, 1] and uses b/n samples per slice, where b is the total number of samples our algorithm makes, analogous to the sketch length b in [23]. |
| Researcher Affiliation | Collaboration | Zhao Song David P. Woodruff Huan Zhang Dept. of Computer Science, University of Texas, Austin, USA IBM Almaden Research Center, San Jose, USA Dept. of Electrical and Computer Engineering, University of California, Davis, USA zhaos@utexas.edu, dpwoodru@us.ibm.com, ecezhang@ucdavis.edu |
| Pseudocode | Yes | Algorithm 1 Subroutine for approximate tensor contraction T(I, v, w) |
| Open Source Code | Yes | We refer the reader to our Git Hub repository 2 for our code and full results. |
| Open Datasets | Yes | We used the two same datasets as the previous work [23]: Wiki and Enron, as well as four additional real-life datasets. We refer the reader to our Git Hub repository 2 for our code and full results. |
| Dataset Splits | No | The paper does not provide specific dataset split information (exact percentages, sample counts, citations to predefined splits, or detailed splitting methodology) for training, validation, or test sets. |
| Hardware Specification | Yes | We ran our experiments on an i7-5820k CPU with 64 GB of memory in singlethreaded mode. |
| Software Dependencies | No | The paper mentions 'Our implementation shares the same code base 1 as the sketching-based robust tensor power method proposed in [23].' and links to a `.zip` file, but does not provide specific software dependencies with version numbers (e.g., Python 3.8, PyTorch 1.9). |
| Experiment Setup | Yes | Then σ controls the noise-to-signal ratio and we kept it as 0.01 in all our synthetic tensors. ... We also set k = 100 when generating tensors... We generated tensors with different dimensions: n = 200, 400, 600, 800, 1000, 1200. |