Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].

Minimax Rates in Permutation Estimation for Feature Matching

Authors: Olivier Collier, Arnak S. Dalalyan

JMLR 2016 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We also discuss the computational aspects of the estimators and provide empirical evidence of their consistency on synthetic data. We carried out a small experimental evaluation that confirms that in the heteroscedastic setting the LSL estimator is as good as the LSNS (pseudo-) estimator and that they outperform the two other estimators: the greedy estimator and the least sum of squares. We have implemented all the procedures in Matlab and carried out numerical experiments on synthetic data.
Researcher Affiliation Academia Olivier Collier EMAIL Imagine LIGM Université Paris EST Marne-la-Vallée, FRANCE; Arnak S. Dalalyan EMAIL Laboratoire de Statistique ENSAE CREST Malakoff, FRANCE
Pseudocode No The paper describes the estimation procedures using mathematical formulas (equations 8-12) and textual explanations, but it does not include any explicitly labeled pseudocode blocks or algorithms in a structured, code-like format.
Open Source Code No The paper states, "We have implemented all the procedures in Matlab and carried out numerical experiments on synthetic data," indicating that code was written for the experiments. However, it does not provide any specific links to a code repository, an explicit statement of code release, or mention of code in supplementary materials.
Open Datasets No We have implemented all the procedures in Matlab and carried out numerical experiments on synthetic data. We chose n = d = 200 and randomly generated a n d matrix θ with i.i.d. entries uniformly distributed on [0, τ], with several values of τ varying between 1.4 and 3.5. Then, we randomly chose a permutation π (uniformly from Sn) and generated the sets {Xi} and {X# i } according to (2) with σi = σ# i = 1.
Dataset Splits No The paper uses synthetic data generated for each trial (e.g., "averaged over 500 independent trials"). It describes the parameters for generating this data but does not mention partitioning a fixed dataset into training, validation, or test sets in the conventional sense of dataset splits.
Hardware Specification No for a problem with n = 500 features, it takes about six seconds to compute a solution to (17) on a standard PC.
Software Dependencies Yes To simplify, we have used the general-purpose solver Se Du Mi (Sturm, 1999) for solving linear programs. Jos F. Sturm. Using Se Du Mi 1.02, a MATLAB toolbox for optimization over symmetric cones. Optim. Methods Softw., 11/12(1-4):625 653, 1999.
Experiment Setup Yes We chose n = d = 200 and randomly generated a n d matrix θ with i.i.d. entries uniformly distributed on [0, τ], with several values of τ varying between 1.4 and 3.5. Then, we randomly chose a permutation π (uniformly from Sn) and generated the sets {Xi} and {X# i } according to (2) with σi = σ# i = 1. The result, averaged over 500 independent trials.