Briola, Antonio and Aste, Tomaso (2023) Topological feature selection. Proceedings of Machine Learning Research, 221. 534 - 556. ISSN 1938-7228
Text (Topological feature selection)
- Published Version
Download (371kB) |
Abstract
In this paper, we introduce a novel unsupervised, graph-based filter feature selection technique which exploits the power of topologically constrained network representations. We model dependency structures among features using a family of chordal graphs (i.e. the Triangulated Maximally Filtered Graph), and we maximise the likelihood of features’ relevance by studying their relative position inside the network. Such an approach presents three aspects that are particularly satisfactory compared to its alternatives: (i) it is highly tunable and easily adaptable to the nature of input data; (ii) it is fully explainable, maintaining, at the same time, a remarkable level of simplicity; (iii) it is computationally cheap. We test our algorithm on 16 benchmark datasets from different application domains showing that it outperforms or matches the current state-of-the-art under heterogeneous evaluation conditions. The code and the data to reproduce all the results presented in the current research work are available at https://github.com/FinancialComputingUCL/Topological_Feature_Selection.
Item Type: | Article |
---|---|
Official URL: | https://proceedings.mlr.press/v221/ |
Additional Information: | © 2023 The Author(s) |
Divisions: | Systemic Risk Centre |
Subjects: | Q Science > Q Science (General) |
Date Deposited: | 15 Dec 2023 10:48 |
Last Modified: | 16 Sep 2024 16:04 |
URI: | http://eprints.lse.ac.uk/id/eprint/121064 |
Actions (login required)
View Item |