Euclidean distance compression via deep random features
Authors: Brett Leroux, Luis Rademacher
NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We validate the method with experiments, including an application to nearest neighbor search. and 4 Experiments The main goal of our experiments is to validate the idea of composing multiple random feature maps. |
| Researcher Affiliation | Academia | Brett Leroux Department of Mathematics University of California, Davis Davis, CA 95616 lerouxbew@gmail.com Luis Rademacher Department of Mathematics University of California, Davis Davis, CA 95616 lrademac@ucdavis.edu |
| Pseudocode | No | The paper describes the construction of the maps φℓ using mathematical definitions and textual descriptions, but it does not include any explicitly labeled pseudocode or algorithm blocks. |
| Open Source Code | Yes | Justification: We included a link to a github repository containing the code used for our experiments in the supplementary material. |
| Open Datasets | Yes | Real data. We perform a similar experiment with the RCV1 dataset [20]. and [20] D. D. Lewis, Y. Yang, T. G. Rose, and F. Li. RCV1: A new benchmark collection for text categorization research. J. Mach. Learn. Res., 5:361 397, 2004. |
| Dataset Splits | No | The paper states, 'We only consider the first 23149 samples which have been previously designated as the training set' for the RCV1 dataset, but it does not provide explicit training/validation/test split percentages or sample counts for all experiments. |
| Hardware Specification | No | The NeurIPS checklist states: 'We ran experiments on a standard laptop and do not think it is necessary to report compute time because we are not including code in the paper.' This does not provide specific hardware details like GPU/CPU models or memory. |
| Software Dependencies | No | The paper does not provide specific software dependencies with version numbers (e.g., library names like PyTorch or TensorFlow, or solver versions). |
| Experiment Setup | Yes | We map x, y by both maps φ1 and φ2 where the output dimension of both maps is 1000 and φ2 first maps to { 1, 1}6000. and We map D by φ1 and φ2 where the output dimension ranges from 23 to 213 and φ2 first maps to the space of dimension six times the output dimension. |