| Full text | |
| Author(s): |
Mantovani, Rafael G.
[1, 2]
;
Rossi, Andre L. D.
[3]
;
Alcobaca, Edesio
[1]
;
Vanschoren, Joaquin
[4]
;
de Carvalho, Andre C. P. L. F.
[1]
Total Authors: 5
|
| Affiliation: | [1] Univ Sao Paulo, Inst Math & Comp Sci, Sao Carlos, SP - Brazil
[2] Fed Technol Univ, Campus Apucarana, R Marcilio Dias 635, BR-86812460 Apucarana, PR - Brazil
[3] Univ Estadual Paulista, Campus Itapeva, Sao Paulo - Brazil
[4] Eindhoven Univ Technol, Eindhoven - Netherlands
Total Affiliations: 4
|
| Document type: | Journal article |
| Source: | INFORMATION SCIENCES; v. 501, p. 193-221, OCT 2019. |
| Web of Science Citations: | 0 |
| Abstract | |
For many machine learning algorithms, predictive performance is critically affected by the hyperparameter values used to train them. However, tuning these hyperparameters can come at a high computational cost, especially on larger datasets, while the tuned settings do not always significantly outperform the default values. This paper proposes a recommender system based on meta-learning to identify exactly when it is better to use default values and when to tune hyperparameters for each new dataset. Besides, an in-depth analysis is performed to understand what they take into account for their decisions, providing useful insights. An extensive analysis of different categories of meta-features, meta-learners, and setups across 156 datasets is performed. Results show that it is possible to accurately predict when tuning will significantly improve the performance of the induced models. The proposed system reduces the time spent on optimization processes, without reducing the predictive performance of the induced models (when compared with the ones obtained using tuned hyperparameters). We also explain the decision-making process of the meta-learners in terms of linear separability-based hypotheses. Although this analysis is focused on the tuning of Support Vector Machines, it can also be applied to other algorithms, as shown in experiments performed with decision trees. (C) 2019 Published by Elsevier Inc. (AU) | |
| FAPESP's process: | 12/23114-9 - Use of meta-learning for parameter tuning for classification problems |
| Grantee: | Rafael Gomes Mantovani |
| Support Opportunities: | Scholarships in Brazil - Doctorate |
| FAPESP's process: | 15/03986-0 - Use of Meta-learning to improve deep learning algorithms in classification problems |
| Grantee: | Rafael Gomes Mantovani |
| Support Opportunities: | Scholarships abroad - Research Internship - Doctorate |
| FAPESP's process: | 18/14819-5 - Automated machine learning: learning to learn |
| Grantee: | Edesio Pinto de Souza Alcobaça Neto |
| Support Opportunities: | Scholarships in Brazil - Doctorate (Direct) |