Domain Adaptation via Transfer Component Analysis

Abstract

Domain adaptation solves a learning problem in a target domain by utilizing the training data in a different but related source domain. Intuitively, discovering a good feature representation across domains is crucial. In this paper, we propose to find such a representation through a new learning method, transfer component analysis (TCA), for domain adaptation. TCA tries to learn some transfer components across domains in a Reproducing Kernel Hilbert Space (RKHS) using Maximum Mean Discrepancy (MMD). In the subspace spanned by these transfer components, data distributions in different domains are close to each other. As a result, with the new representations in this subspace, we can apply standard machine learning methods to train classifiers or regression models in the source domain for use in the target domain. The main contribution of our work is that we propose a novel feature representation in which to perform domain adaptation via a new parametric kernel using feature extraction methods, which can dramatically minimize the distance between domain distributions by projecting data onto the learned transfer components. Furthermore, our approach can handle large datsets and naturally lead to out-of-sample generalization. The effectiveness and efficiency of our approach in are verified by experiments on two real-world applications: cross-domain indoor WiFi localization and cross-domain text classification.

Exploiting kernel-based feature weighting and instance clustering to transfer knowledge across domains

“…Transfer learning and domain adaptation [3,4] are challenging research areas in recent years and they have been comprehensively studied from various perspectives, including natural language processing [10,14], statistics and machine learning [12,15], and recently computer vision [16][17][18][19]. Pan et al [4] presented a complete survey of cross-domain learning methods, and discussed the different applications of transfer learning.…”

Adapting pedestrian detectors to new domains: A comprehensive review

“…Classifiers trained on different source domains iteratively train and improve each other by teaching, in each iteration, the most confident detections of one classifier to the other classifier(s). The algorithm makes use of Transfer Component Analysis [41] in order to reduce the effects of domain shifts between the datasets. As with other iterative self-training algorithms, the algorithm requires setting the threshold for selecting confident detections.…”

Learning from the Mistakes of Others: Matching Errors in Cross-Dataset Learning

“…The empirical Maximum Mean Discrepancy (MMD) [12] is employed as the nonparametric metric of difference between two distributions. In the domain adaptation setting, Pan et al [23] used the MMD metric to project data from a target domain X := {x 1 , . .…”