Passer à la navigation principale Passer à la recherche Passer au contenu principal

Fully-Dynamic Approximate Decision Trees With Worst-Case Update Time Guarantees

  • Marco Bressan
  • , Mauro Sozio
  • University of Milano
  • Institut Polytechnique de Paris

Résultats de recherche: Contribution à un journalArticle de conférenceRevue par des pairs

Résumé

We study the problem of maintaining a decision tree in the fully-dynamic setting, where the dataset is updated by an adversarial sequence of insertions and deletions. We present the first algorithm with strong guarantees on both the quality of the tree and the worst-case update time (the maximum number of operations the algorithm performs between two dataset updates). For instance, we can maintain a tree where each node has Gini gain within β of the optimum, while guaranteeing an update time O(dβ-3 log4 n), where d is the number of features and n the maximum size of the dataset. This is optimal up to polylogarithmic factors, as any dynamic algorithm must have update time in Ω(d). Similar guarantees hold for the variance and information gain, for classification and regression, as well as for boosted trees. This shows that many popular decision trees such as ID3 or C4.5 can be efficiently made dynamic, answering an open question of Bressan, Damay, and Sozio (2023).

langue originaleAnglais
Pages (de - à)4517-4541
Nombre de pages25
journalProceedings of Machine Learning Research
Volume235
étatPublié - 1 janv. 2024
Modification externeOui
Evénement41st International Conference on Machine Learning, ICML 2024 - Vienna, Autriche
Durée: 21 juil. 202427 juil. 2024

Empreinte digitale

Examiner les sujets de recherche de « Fully-Dynamic Approximate Decision Trees With Worst-Case Update Time Guarantees ». Ensemble, ils forment une empreinte digitale unique.

Contient cette citation