« Projekte
Sie verwenden einen sehr veralteten Browser und können Funktionen dieser Seite nur sehr eingeschränkt nutzen. Bitte aktualisieren Sie Ihren Browser. http://www.browser-update.org/de/update.html
Online adaptation to the parameters of the problem in active learning
Finanzierung:
Deutsche Forschungsgemeinschaft (DFG) ;
We consider the problem of active learning in the setting of classification and optimisation, which are baselines problems in applied mathematics and machine learning. The problem of adaptivity (to unknown distributional parameters) has remained opened in many contexts (e.g. smooth decision boundary for classification, or optimisation of the cumulative regret). While some recent advances on this problems established adaptive rates in some contexts, adaptivity in most of the real world setting has so far remained elusive. In this project, we plan to investigate the problem of adapting to the unknown parameters of the problem (e.g. smoothness, margin assumptions, measure of the near optimal points, etc), and intend to develop minimax rates on the learning efficiency with respect to an oracle learner.

Schlagworte

active learning, machine learning
Kontakt

weitere Projekte

Die Daten werden geladen ...