Identity testing for Mallows model
Authors: Róbert Busa-Fekete, Dimitris Fotakis, Balazs Szorenyi, Emmanouil Zampetakis
NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We shall present synthetic experiments to assess the performance asymptotic and non-asymptotic tests. Each result are computed based on 100 repetitions. |
| Researcher Affiliation | Collaboration | Róbert Busa-Fekete Google Research, New York, USA... Dimitris Fotakis National Technical University of Athens, Greece... Balázs Szörényi Yahoo! Research New York, USA... Manolis Zampetakis Department of Statistics University of California, Berkeley, USA |
| Pseudocode | Yes | Algorithm 1 Non-asymptotic test for spread parameter based on UMPU. Algorithm 2 Non-asymptotic test for spread parameter. |
| Open Source Code | No | Did you include the code, data, and instructions needed to reproduce the main experimental results (either in the supplemental material or as a URL)? [No] We used synthetic data. |
| Open Datasets | No | Did you include the code, data, and instructions needed to reproduce the main experimental results (either in the supplemental material or as a URL)? [No] We used synthetic data. |
| Dataset Splits | No | The paper states "We shall present synthetic experiments" and mentions sample sizes like "n = {10, 50} random rankings" and "single sample", but it does not specify explicit training/validation/test dataset splits (e.g., percentages or sample counts) for reproducibility. |
| Hardware Specification | No | Did you include the total amount of compute and the type of resources used (e.g., type of GPUs, internal cluster, or cloud provider)? [No] |
| Software Dependencies | No | The paper does not specify any software dependencies with version numbers required to replicate the experiments. |
| Experiment Setup | Yes | The goal of the first set of experiments is to compare the power of the non-asymptotic test fn defined in Theorem 5.1 based on exact computation of the distribution of the sufficient statistic and its χ2 approximation. Figure 2 shows their power for various null and alternative hypothesis. We picked the central ranking to be the identity ranking, since it has no impact on the results. One can see that the approximation slightly deteriorates the power of fn for small sample size, but the difference is marginal for n = 50. We run Algorithm 2 with log (1/δ) /(2m"2 0). |