A Distance-Weighted Class-Homogeneous Neighbourhood Ratio for Algorithm Selection

Haofei Chen, Ya Liu, Japnit Kaur Ahuja, Daren Ler
Proceedings of The 12th Asian Conference on Machine Learning, PMLR 129:1-16, 2020.

Abstract

In this paper, we introduce a new form of meta-feature that is based on a distance-weighted class-homogeneous neighbourhood ratio to facilitate algorithm selection. We show that these new meta-features, while exhibiting a cost advantage, achieve a comparable, and in some cases, higher performance than conventional meta-features. These results were obtained via experiments conducted over artificial datasets and real-world datasets from the UCI repository. We further redefine the algorithm selection problem by advocating that accuracy should be calculated based on the assumption that the population of datasets is uniformly distributed. Finally, in this paper, we provide a new perspective on landmarkers, such that a landmarker corresponds to a tuple (algorithm, metric), and propose the idea of a new family of meta-features.

Cite this Paper

BibTeX
@InProceedings{pmlr-v129-chen20a, title = {A Distance-Weighted Class-Homogeneous Neighbourhood Ratio for Algorithm Selection}, author = {Chen, Haofei and Liu, Ya and Ahuja, Japnit Kaur and Ler, Daren}, booktitle = {Proceedings of The 12th Asian Conference on Machine Learning}, pages = {1--16}, year = {2020}, editor = {Pan, Sinno Jialin and Sugiyama, Masashi}, volume = {129}, series = {Proceedings of Machine Learning Research}, month = {18--20 Nov}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v129/chen20a/chen20a.pdf}, url = {https://proceedings.mlr.press/v129/chen20a.html}, abstract = {In this paper, we introduce a new form of meta-feature that is based on a distance-weighted class-homogeneous neighbourhood ratio to facilitate algorithm selection. We show that these new meta-features, while exhibiting a cost advantage, achieve a comparable, and in some cases, higher performance than conventional meta-features. These results were obtained via experiments conducted over artificial datasets and real-world datasets from the UCI repository. We further redefine the algorithm selection problem by advocating that accuracy should be calculated based on the assumption that the population of datasets is uniformly distributed. Finally, in this paper, we provide a new perspective on landmarkers, such that a landmarker corresponds to a tuple (algorithm, metric), and propose the idea of a new family of meta-features.} }
Endnote
%0 Conference Paper %T A Distance-Weighted Class-Homogeneous Neighbourhood Ratio for Algorithm Selection %A Haofei Chen %A Ya Liu %A Japnit Kaur Ahuja %A Daren Ler %B Proceedings of The 12th Asian Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2020 %E Sinno Jialin Pan %E Masashi Sugiyama %F pmlr-v129-chen20a %I PMLR %P 1--16 %U https://proceedings.mlr.press/v129/chen20a.html %V 129 %X In this paper, we introduce a new form of meta-feature that is based on a distance-weighted class-homogeneous neighbourhood ratio to facilitate algorithm selection. We show that these new meta-features, while exhibiting a cost advantage, achieve a comparable, and in some cases, higher performance than conventional meta-features. These results were obtained via experiments conducted over artificial datasets and real-world datasets from the UCI repository. We further redefine the algorithm selection problem by advocating that accuracy should be calculated based on the assumption that the population of datasets is uniformly distributed. Finally, in this paper, we provide a new perspective on landmarkers, such that a landmarker corresponds to a tuple (algorithm, metric), and propose the idea of a new family of meta-features.
APA
Chen, H., Liu, Y., Ahuja, J.K. & Ler, D.. (2020). A Distance-Weighted Class-Homogeneous Neighbourhood Ratio for Algorithm Selection. Proceedings of The 12th Asian Conference on Machine Learning, in Proceedings of Machine Learning Research 129:1-16 Available from https://proceedings.mlr.press/v129/chen20a.html.

Related Material