Fast Min-ϵ Segmented Regression using Constant-Time Segment Merging.
| Author: | Lößer, A., Schlecht, M., Schintke, F., Witzke, J., & Scheuermann, B. |
| Published in: | Proceedings of Machine Learning Research (ICML ’25 paper), 267, 40312–40327 |
| Year: | 2025 |
| Type: | Academic articles |
Segmented regression is a statistical method that approximates a function f by a piecewise function f using noisy data samples. Min-ϵ approaches aim to reduce the regression function’s mean squared error (MSE) for a given number of kseg- ments. An optimal solution for min-ϵsegmented regression is found in O(n2) time (Bai & Perron, 1998; Yamamoto & Perron, 2013) for nsamples. For large datasets, current heuristics improve time complexity to O(nlog n) (Acharya et al., 2016) but can result in large errors, especially when ex- actly ksegments are used. We present a method for min-ϵsegmented regression that combines the scalability of top existing heuristic solutions with a statistical efficiency similar to the optimal so- lution. This is achieved by using a new method to merge an initial set of segments using precom- puted matrices from samples, allowing both merg- ing and error calculation in constant time. Our approach, using the same samples and parameter k, produces segments with up to 1,000×lower MSE compared to Acharya et al. (2016) in about 100×less runtime on datasets over 104 samples.
| Visit publication |

Connected HIIG researchers
Björn Scheuermann, Prof. Dr.
- Open Access
- Peer Reviewed
