Boosting for Comparison-Based Learning

Authors: Michael Perrot, Ulrike von Luxburg

IJCAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental we empirically show that our method is both competitive with state of the art approaches and resistant to noise. We propose an empirical evaluation of Triplet Boost. We consider six datasets of varying scales and four baselines.
Researcher Affiliation Academia 1Max-Planck-Institute for Intelligent Systems, T ubingen, Germany 2University of T ubingen, Department of Computer Science, T ubingen, Germany
Pseudocode Yes Algorithm 1 Triplet Boost: boosting with triplet classifiers
Open Source Code No The paper mentions using "sk-learn" and "our own implementation" for some components, but it does not provide any statement or link indicating that the source code for Triplet Boost itself is publicly available.
Open Datasets Yes We consider six datasets: Iris, Moons, Gisette, Cod-rna, MNIST, and k MNIST. As a proof of concept we considered the 1m movielens dataset [Harper and Konstan, 2016].
Dataset Splits No The paper mentions 'training examples' and 'test accuracy' but does not specify explicit percentages or counts for training, validation, and test dataset splits, nor does it refer to standard predefined splits with citations for these datasets.
Hardware Specification Yes The results were obtained on a single core @3.40GHz.
Software Dependencies No The paper mentions using "sk-learn" but does not provide a specific version number for it or any other software dependency relevant to reproducibility.
Experiment Setup Yes For t STE we used the implementation distributed on the authors website and we set the embedding dimension to the original dimension of the data. [...] For Comp Tree we used our own implementation and the leaf size of the comparison tree is set to 1 as this is the only value for which this method can handle noise. [...] The number of boosting iterations for SAMME is set to 103. Finally for Triplet Boost we set the number of iterations to 106.