We introduce a framework based on bilevel programming that unifies gradient-based hyperparameter optimization and meta-learning. We show that an approximate version of the bilevel problem can be solved by taking into explicit account the optimization dynamics for the inner objective. Depending on the specific setting, the outer variables take either the meaning of hyperparameters in a supervised learning problem or parameters of a meta-learner. We provide sufficient conditions under which solutions of the approximate problem converge to those of the exact problem. We instantiate our approach for meta-learning in the case of deep learning where representation layers are treated as hyperparameters shared across a set of training episodes. In experiments, we confirm our theoretical findings, present encouraging results for few-shot learning and contrast the bilevel approach against classical approaches for learning-to-learn.
Bilevel Programming for Hyperparameter Optimization and Meta-Learning / Franceschi, Luca; Frasconi, Paolo; Salzo, Saverio; Grazzi, Riccardo; Pontil, Massimiliano. - ELETTRONICO. - 80:(2018), pp. 2537-2548. (Intervento presentato al convegno International Conference on Machine Learning tenutosi a Stockholm nel 10-15 July 2018).
Bilevel Programming for Hyperparameter Optimization and Meta-Learning
Frasconi, Paolo;
2018
Abstract
We introduce a framework based on bilevel programming that unifies gradient-based hyperparameter optimization and meta-learning. We show that an approximate version of the bilevel problem can be solved by taking into explicit account the optimization dynamics for the inner objective. Depending on the specific setting, the outer variables take either the meaning of hyperparameters in a supervised learning problem or parameters of a meta-learner. We provide sufficient conditions under which solutions of the approximate problem converge to those of the exact problem. We instantiate our approach for meta-learning in the case of deep learning where representation layers are treated as hyperparameters shared across a set of training episodes. In experiments, we confirm our theoretical findings, present encouraging results for few-shot learning and contrast the bilevel approach against classical approaches for learning-to-learn.File | Dimensione | Formato | |
---|---|---|---|
franceschi18a.pdf
accesso aperto
Descrizione: Articolo principale
Tipologia:
Pdf editoriale (Version of record)
Licenza:
Open Access
Dimensione
1.78 MB
Formato
Adobe PDF
|
1.78 MB | Adobe PDF |
I documenti in FLORE sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.