Transfer Feature Learning With Joint Distribution Adaptation

3142019 Therefore a modified transfer learning algorithm with Joint Distribution Adaptation JDA and Maximum Margin Criterion MMC is put forward in this paper which we call MMC-JDA for short.

Transfer feature learning with joint distribution adaptation. Feature matching and instance reweighting. 212020 The weak form of transfer learning with domain adaptation is to learn a feature transform that simultaneously minimizes the discrepancy between marginal distribution and conditional distribution ie 4 min D P s ϕ X s P t ϕ X t 5 a n d min D Q s Y s ϕ X s Q t Y t ϕ X t where D is the function to evaluate the domain discrepancy. We therefore put forward a novel transfer learning and domain adaptation approach referred to as visual domain adaptation VDA.

Transfer learning aims to learn the labels yt of Dt using the source domain Ds. Specifically JDA aims to jointly adapt both the marginal distribution and conditional distribution in a principled dimensionality reduction procedure and construct new feature representation that is effective and robust for. Transfer learning and domain adaptation are promising solutions to solve the problem that the training set source domain and the test set target domain follow different distributions.

1 Pxs and Pxt2Pysxs. 572016 In this paper we exploit joint transfer learning and domain adaptation to cope with domain shift problem in which the distribution difference is significantly large particularly vision datasets. In this paper we put forward a novel transfer learning approach referred to as Joint Distribution Adaptation JDA.

By Mingsheng Long Jianmin Wang Guiguang Ding et al. Specifi-cally JDA aims to jointly adapt both the marginal distribu-tion and conditional distribution in a principled dimension-ality reduction procedure and construct new feature repre-sentation that is effective and robust for substantial distribu-tion difference. Learn transferable features for adapting models from a source domain to a different target domain.

Transfer feature learning with joint distribution adaptationCProceedings of the IEEE International Conference on Computer Vision. Inspired by the idea of transfer learning that leverages the knowledge learnt from rich labeled data in source domain to facilitate diagnosing a new but similar target task a new intelligent fault diagnosis framework ie deep transfer network DTN which generalizes deep learning model to domain adaptation scenario is proposed in this paper. Specifically JDA aims to jointly adapt both the marginal distribution and conditional distribution in a principled dimensionality reduction procedure and construct new feature representation that is effective and robust for substantial distribution difference.

Specifically JDA aims to jointly adapt both the marginal distribution and conditional distribution in a principled dimensionality reduction procedure and construct new feature representation that is effective and robust for substantial distribution difference. To minimize the discrepancies between. Balanced distribution adaptation solves the transfer learn-ing problem by adaptively minimizing the marginal and conditional distribution discrepancy between domains and handle the class imbalance problem ie.