Constraining the parameters of high-dimensional models with active learning
Abstract Constraining the parameters of physical models with $$>5-10$$ >5-10 parameters is a widespread problem in fields like particle physics and astronomy. The generation of data to explore this parameter space often requires large amounts of computational resources. The commonly used solut...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
SpringerOpen
2019-11-01
|
| Series: | European Physical Journal C: Particles and Fields |
| Online Access: | http://link.springer.com/article/10.1140/epjc/s10052-019-7437-5 |
| _version_ | 1818459023556476928 |
|---|---|
| author | Sascha Caron Tom Heskes Sydney Otten Bob Stienen |
| author_facet | Sascha Caron Tom Heskes Sydney Otten Bob Stienen |
| author_sort | Sascha Caron |
| collection | DOAJ |
| description | Abstract Constraining the parameters of physical models with $$>5-10$$ >5-10 parameters is a widespread problem in fields like particle physics and astronomy. The generation of data to explore this parameter space often requires large amounts of computational resources. The commonly used solution of reducing the number of relevant physical parameters hampers the generality of the results. In this paper we show that this problem can be alleviated by the use of active learning. We illustrate this with examples from high energy physics, a field where simulations are often expensive and parameter spaces are high-dimensional. We show that the active learning techniques query-by-committee and query-by-dropout-committee allow for the identification of model points in interesting regions of high-dimensional parameter spaces (e.g. around decision boundaries). This makes it possible to constrain model parameters more efficiently than is currently done with the most common sampling algorithms and to train better performing machine learning models on the same amount of data. Code implementing the experiments in this paper can be found on GitHub |
| first_indexed | 2024-12-14T23:07:46Z |
| format | Article |
| id | doaj.art-429c0d17b124447ca206bf462afa3306 |
| institution | Directory Open Access Journal |
| issn | 1434-6044 1434-6052 |
| language | English |
| last_indexed | 2024-12-14T23:07:46Z |
| publishDate | 2019-11-01 |
| publisher | SpringerOpen |
| record_format | Article |
| series | European Physical Journal C: Particles and Fields |
| spelling | doaj.art-429c0d17b124447ca206bf462afa33062022-12-21T22:44:17ZengSpringerOpenEuropean Physical Journal C: Particles and Fields1434-60441434-60522019-11-01791111110.1140/epjc/s10052-019-7437-5Constraining the parameters of high-dimensional models with active learningSascha Caron0Tom Heskes1Sydney Otten2Bob Stienen3Institute for Mathematics, Astro- and Particle Physics IMAPP, Radboud UniversiteitData Science Institute for Computing and Information Sciences (iCIS), Radboud UniversiteitInstitute for Mathematics, Astro- and Particle Physics IMAPP, Radboud UniversiteitInstitute for Mathematics, Astro- and Particle Physics IMAPP, Radboud UniversiteitAbstract Constraining the parameters of physical models with $$>5-10$$ >5-10 parameters is a widespread problem in fields like particle physics and astronomy. The generation of data to explore this parameter space often requires large amounts of computational resources. The commonly used solution of reducing the number of relevant physical parameters hampers the generality of the results. In this paper we show that this problem can be alleviated by the use of active learning. We illustrate this with examples from high energy physics, a field where simulations are often expensive and parameter spaces are high-dimensional. We show that the active learning techniques query-by-committee and query-by-dropout-committee allow for the identification of model points in interesting regions of high-dimensional parameter spaces (e.g. around decision boundaries). This makes it possible to constrain model parameters more efficiently than is currently done with the most common sampling algorithms and to train better performing machine learning models on the same amount of data. Code implementing the experiments in this paper can be found on GitHubhttp://link.springer.com/article/10.1140/epjc/s10052-019-7437-5 |
| spellingShingle | Sascha Caron Tom Heskes Sydney Otten Bob Stienen Constraining the parameters of high-dimensional models with active learning European Physical Journal C: Particles and Fields |
| title | Constraining the parameters of high-dimensional models with active learning |
| title_full | Constraining the parameters of high-dimensional models with active learning |
| title_fullStr | Constraining the parameters of high-dimensional models with active learning |
| title_full_unstemmed | Constraining the parameters of high-dimensional models with active learning |
| title_short | Constraining the parameters of high-dimensional models with active learning |
| title_sort | constraining the parameters of high dimensional models with active learning |
| url | http://link.springer.com/article/10.1140/epjc/s10052-019-7437-5 |
| work_keys_str_mv | AT saschacaron constrainingtheparametersofhighdimensionalmodelswithactivelearning AT tomheskes constrainingtheparametersofhighdimensionalmodelswithactivelearning AT sydneyotten constrainingtheparametersofhighdimensionalmodelswithactivelearning AT bobstienen constrainingtheparametersofhighdimensionalmodelswithactivelearning |