Explicit Inductive Bias for Transfer Learning with Convolutional Networks - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2018

Explicit Inductive Bias for Transfer Learning with Convolutional Networks

Résumé

In inductive transfer learning, fine-tuning pre-trained convolutional networks substantially out-performs training from scratch. When using fine-tuning, the underlying assumption is that the pre-trained model extracts generic features, which are at least partially relevant for solving the target task, but would be difficult to extract from the limited amount of data available on the target task. However, besides the initialization with the pre-trained model and the early stopping, there is no mechanism in fine-tuning for retaining the features learned on the source task. In this paper , we investigate several regularization schemes that explicitly promote the similarity of the final solution with the initial model. We show the benefit of having an explicit inductive bias towards the initial model, and we eventually recommend a simple L 2 penalty with the pre-trained model being a reference as the baseline of penalty for transfer learning tasks.
Fichier principal
Vignette du fichier
for_arxiv.pdf (342.32 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-01843169 , version 1 (18-07-2018)

Identifiants

  • HAL Id : hal-01843169 , version 1

Citer

Xuhong Li, Yves Grandvalet, Franck Davoine. Explicit Inductive Bias for Transfer Learning with Convolutional Networks. 35th International Conference on Machine Learning (ICML 2018), Jul 2018, Stockholm, Sweden. pp.2825-2834. ⟨hal-01843169⟩
216 Consultations
79 Téléchargements

Partager

Gmail Facebook X LinkedIn More