Regressive Virtual Metric Learning

Authors: Michaël Perrot, Amaury Habrard

NeurIPS 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Lastly, we evaluate our approach on several state of the art datasets. and Section 4 is dedicated to an empirical evaluation of our method on several widely used datasets.
Researcher Affiliation Academia Micha el Perrot, and Amaury Habrard Universit e de Lyon, Universit e Jean Monnet de Saint-Etienne, Laboratoire Hubert Curien, CNRS, UMR5516, F-42000, Saint-Etienne, France. {michael.perrot,amaury.habrard}@univ-st-etienne.fr
Pseudocode Yes Algorithm 1: Selecting S from a set of examples S.
Open Source Code Yes The closed-form implementation of RVML is freely available on the authors website.
Open Datasets Yes In this section, we evaluate our approach on 13 different datasets coming from either the UCI [19] repository or used in recent works in metric learning [8, 20, 21].
Dataset Splits Yes For isolet, splice and svmguide1 we have access to a standard training/test partition, for the other datasets we use a 70% training/30% test partition, we perform the experiments on 10 different splits and we average the result. We set our regularization parameter λ with a 5-fold cross validation.
Hardware Specification No The paper does not explicitly describe the specific hardware (e.g., GPU/CPU models, memory) used for running the experiments.
Software Dependencies No The paper mentions algorithms and methods (e.g., 'Sinkhorn-Knopp algorithm', '1-nearest neighbor classifier', 'SCML', 'LMNN') but does not specify software dependencies with version numbers (e.g., Python 3.x, PyTorch 1.x).
Experiment Setup Yes We normalize the examples with respect to the training set by subtracting for each attribute its mean and dividing by 3 times its standard deviation. We set our regularization parameter λ with a 5-fold cross validation. After the metric learning step, we use a 1-nearest neighbor classifier to assess the performance of the metric and report the accuracy obtained. and with the parameter σ fixed as the mean of all pairwise training set Euclidean distances