![[2112.06251] Learning with Subset Stacking](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2112.06251] Learning with Subset Stacking
Summary
The paper introduces a novel regression algorithm called Learning with Subset Stacking (LESS), which effectively learns from heterogeneous input-output relationships by generating subsets around random points in the input space and combining local predictors.
Why It Matters
This research is significant as it addresses the challenges of modeling complex relationships in machine learning, particularly in scenarios where data exhibits heterogeneous behavior. The proposed algorithm shows competitive performance against state-of-the-art methods, potentially enhancing predictive accuracy in various applications.
Key Takeaways
- LESS algorithm generates subsets around random input points to improve regression accuracy.
- Local predictors are trained for each subset, enhancing the model's adaptability.
- The algorithm includes bagging and boosting variants, expanding its applicability.
- LESS demonstrates competitive performance against existing state-of-the-art methods.
- This approach could be beneficial for complex datasets with heterogeneous relationships.
Computer Science > Machine Learning arXiv:2112.06251 (cs) [Submitted on 12 Dec 2021 (v1), last revised 15 Feb 2026 (this version, v4)] Title:Learning with Subset Stacking Authors:Ş. İlker Birbil, Sinan Yıldırım, Samet Çopur, M. Hakan Akyüz View a PDF of the paper titled Learning with Subset Stacking, by \c{S}. \.Ilker Birbil and 3 other authors View PDF HTML (experimental) Abstract:We propose a new regression algorithm that learns from a set of input-output pairs. Our algorithm is designed for populations where the relation between the input variables and the output variable exhibits a heterogeneous behavior across the predictor space. The algorithm starts with generating subsets that are concentrated around random points in the input space. This is followed by training a local predictor for each subset. Those predictors are then combined in a novel way to yield an overall predictor. We call this algorithm "LEarning with Subset Stacking" or LESS, due to its resemblance to the method of stacking regressors. We offer bagging and boosting variants of LESS and test against the state-of-the-art methods on several datasets. Our comparison shows that LESS is highly competitive. Comments: Subjects: Machine Learning (cs.LG); Machine Learning (stat.ML) Cite as: arXiv:2112.06251 [cs.LG] (or arXiv:2112.06251v4 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2112.06251 Focus to learn more arXiv-issued DOI via DataCite Submission history From: Ilker Birbil [view email] [v1]...
