Learning With Augmented Features for Supervised and Semi-Supervised Heterogeneous Domain Adaptation

By Wen Li, Lixin Duan, Dong Xu, Ivor Wai-Hung Tsang. 2014

428 citations recorded by the source corpus.

In this paper, we study the heterogeneous domain adaptation (HDA) problem, in which the data from the source domain and the target domain are represented by heterogeneous features with different dimensions. By introducing two different projection matrices, we first transform the data from two domains into a common subspace such that the similarity between samples across different domains can be measured. We then propose a new feature mapping function for each domain, which augments the transformed samples with their original features and zeros. Existing supervised learning methods (e.g., SVM and SVR) can be readily employed by incorporating our newly proposed augmented feature representations for supervised HDA. As a showcase, we propose a novel method called Heterogeneous Feature Augmentation (HFA) based on SVM. We show that the proposed formulation can be equivalently derived as a standard Multiple Kernel Learning (MKL) problem, which is convex and thus the global solution can be guaranteed. To additionally utilize the unlabeled data in the target domain, we further propose the semi-supervised HFA (SHFA) which can simultaneously learn the target classifier as well as infer the labels of unlabeled target samples. Comprehensive experiments on three different applications clearly demonstrate that our SHFA and HFA outperform the existing HDA methods.

Subfields: Artificial Intelligence, Cancer Research

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Cancer Research

Citations in the corpus

Unsupervised Domain Adaptation with Residual Transfer Networks (2016, background): “icitly taking the cross-domain discrepancy into account. Transfer learning is to mitigate the burden of manual labeling for machine learning (Pan et al., 2011; Duan et al., 2012a; Zhang et al., 2013; Li et al., 2014; Wang & Schneider, 2014), computer vision (Saenko et al., 2010; Gopalan et al., ”
Domain Adaptation for Visual Applications: A Comprehensive Survey (2017, positive): “in [134], where each projection matrix is decomposed into a linear combination of a set of rank-one positive semi-definite matrices allowing the usage of Multiple Kernel Learning to find the solution.”
Learning Compatible Embeddings (2021, methodology): “Feature augmentation based methods [7, 21, 10, 22] transfer the original features by feature replication to augment datasets for target domain.”
Instilling Social to Physical: Co-Regularized Heterogeneous Transfer Learning (2016, methodology): “While this work tackled the clustering task, later a series of studies (Dai et al. 2008; Duan et al. 2012; Li et al. 2014; Shi et al. 2013; Wang and Mahadevan 2011; Zhu et al. 2011) focused on classification.”
Recognizing RGB Images by Learning from RGB-D Data (2014, methodology): “Moreover, the heterogeneous domain adaptation (HDA) methods [22, 24] also cannot be used because the labeled samples in the target domain are required in these HDA methods.”
Transfer Neural Trees for Heterogeneous Domain Adaptation (2016, methodology): “DS → DT SVMt NNt MMDT [18] HFA [21] SHFR [19] SCP [23] TNT tag → image 63.”
Multi-view Domain Generalization for Visual Recognition (2015, background): “Recently, a large number of domain adaptation methods [20, 19, 1, 17, 6, 24, 14, 13, 15, 26, 27] have been proposed to explicitly cope with the data distribution mismatch between the training samples from the source domain and the testing samples from the target domain.”
Domain adaptive sparse representation-based classification (2015, background): “This work was extended for the heterogeneous data in [3].”
An introduction to domain adaptation and transfer learning (2018, methodology): “Later approaches are more nonlinear through the use of kernelization or by employing ”co-training” approaches [40,136].”
Visual Domain Adaptation: A survey of recent advances (2015, background): “( ) P i H”
Person Re-Identification by Camera Correlation Aware Feature Augmentation (2017, methodology): “Our model formulation is also related to the zero padding technique [71], [72] and feature data augmentation methods in [67], [69].”
RGB-Infrared Cross-Modality Person Re-identification (2017, positive): “, HFA [18], CRAFT [3], zero-padding in feature level is applied and proved to be effective.”
Heterogeneous Graph Attention Network for Unsupervised Multiple-Target Domain Adaptation (2020, background): “Inspired by this, we adopt following semantic transfer objective for UDA: where C mS and C mT k are centroid for each class in feature space, Φ is the measure function with the squared Euclidean distance, and we use 2M centroids.”
Exploiting Low-Rank Structure from Latent Domains for Domain Generalization (2014, methodology): “Domain adaptation techniques, which aim to reduce the domain distribution mismatch when the training and testing samples come from different domains, have been successfully used for a broad range of vision applications such as object recognition and video event recognition [23,17,16,10,11,12,6,7,24]”
Transferable Representation Learning with Deep Adaptation Networks (2019, background): “Transfer learning mitigates the burden of manual labeling in machine learning [3], [5], [6], [7], [24], [25], computer vision [4], [20], [26], [27], [28], [29] and natural language processing [30], [31].”
Heterogeneous Few-Shot Model Rectification With Semantic Mapping (2020, methodology): “Three more heterogeneous transfer learning methods are also compared, namely, HFA [61], OTL [62], and MMDT [63].”