- Research Article
45
- 10.1016/j.patcog.2021.108238
Universal multi-Source domain adaptation for image classification
- Aug 14, 2021
- Pattern Recognition
- Yueming Yin + 3 more +3
Universal multi-Source domain adaptation for image classification
Unsupervised domain adaptation provides a variety of methods to leverage the previously gained knowledge from a labeled source domain to help complete a task from a similar unlabeled target domain. Many existing methods focus on transferring knowledge across single source and single target domains, while few studies deal with multi-source domain adaptation, which is more realistic and challengeable. Existing multi-source domain adaptation methods rarely consider the uncertainty of the transformed knowledge resulting from limited information in target domain. A fuzzy system allows imprecision and ambiguity within transfer, thus it can deal with problems with uncertainty. This work proposes a multi-source domain adaptation method with fuzzy-rule based deep neural networks (MDAFuz). The proposed method first extracts multi-view adapted features and pre-trains source classifiers. Using the learned features and classifiers, training samples are then split into multiple clusters, hence fuzzy rules can be built to learn new classifiers. At the same time, the cluster discriminator is trained to define the membership. Finally, by measuring the similarities among source and target domains using the pseudo target labels and a domain discriminator, the target task is completed by combining all source classifiers with regard to the learned weights. The experiment results on real-world visual datasets show the superiority of the proposed method.
Universal multi-Source domain adaptation for image classification
Universal multi-Source domain adaptation for image classification
Multi-source domain adaptation with graph embedding and adaptive label prediction
Multi-source domain adaptation with graph embedding and adaptive label prediction
Multi-Source Contribution Learning for Domain Adaptation.
Transfer learning becomes an attractive technology to tackle a task from a target domain by leveraging previously acquired knowledge from a similar domain (source domain). Many existing transfer learning methods focus on learning one discriminator with single-source domain. Sometimes, knowledge from single-source domain might not be enough for predicting the target task. Thus, multiple source domains carrying richer transferable information are considered to complete the target task. Although there are some previous studies dealing with multi-source domain adaptation, these methods commonly combine source predictions by averaging source performances. Different source domains contain different transferable information; they may contribute differently to a target domain compared with each other. Hence, the source contribution should be taken into account when predicting a target task. In this article, we propose a novel multi-source contribution learning method for domain adaptation (MSCLDA). As proposed, the similarities and diversities of domains are learned simultaneously by extracting multi-view features. One view represents common features (similarities) among all domains. Other views represent different characteristics (diversities) in a target domain; each characteristic is expressed by features extracted in a source domain. Then multi-level distribution matching is employed to improve the transferability of latent features, aiming to reduce misclassification of boundary samples by maximizing discrepancy between different classes and minimizing discrepancy between the same classes. Concurrently, when completing a target task by combining source predictions, instead of averaging source predictions or weighting sources using normalized similarities, the original weights learned by normalizing similarities between source and target domains are adjusted using pseudo target labels to increase the disparities of weight values, which is desired to improve the performance of the final target predictor if the predictions of sources exist significant difference. Experiments on real-world visual data sets demonstrate the superiorities of our proposed method.
Read moreUnsupervised Adversarial Domain Adaptation Network for Semantic Segmentation
With the rapid development of deep learning technology, semantic segmentation methods have been widely used in remote sensing data. A pretrained semantic segmentation model usually cannot perform well when the testing images (target domain) have an obvious difference from the training data set (source domain), while a large enough labeled data set is almost impossible to be acquired for each scenario. Unsupervised domain adaptation (DA) techniques aim to transfer knowledge learned from the source domain to a totally unlabeled target domain. By reducing the domain shift, DA methods have shown the ability to improve the classification accuracy for the target domain. Hence, in this letter, we propose an unsupervised adversarial DA network that converts deep features into 2-D feature curves and reduces the discrepancy between curves from the source domain and curves from the target domain based on a conditional generative adversarial networks (cGANs) model. Our proposed DA network is able to improve the semantic labeling accuracy when we apply a pretrained semantic segmentation model to the target domain. To test the effectiveness of the proposed method, experiments are conducted on the International Society for Photogrammetry and Remote Sensing (ISPRS) 2-D Semantic Labeling data set. Results show that our proposed network is able to stably improve overall accuracy not only when the source and target domains are from the same city but with different building styles but also when the source and target domains are from different cities and acquired by different sensors. By comparing with a few state-of-the-art DA methods, we demonstrate that our proposed method achieves the best cross-domain semantic segmentation performance.
Read moreEDTM: Efficient Domain Transition for Multi-Source Domain Adaptation
Domain adaptation aims to reduce the distribution gap between the training data (source domain) and the target data. This enables effective predictions even for domains not seen during training. However, most conventional domain adaptation methods assume a single source domain, making them less suitable for modern deep learning settings that rely on diverse and large-scale datasets. To address this limitation, recent research has focused on Multi-Source Domain Adaptation (MSDA), which aims to learn effectively from multiple source domains. In this paper, we propose Efficient Domain Transition for Multi-source (EDTM), a novel and efficient framework designed to tackle two major challenges in existing MSDA approaches: (1) integrating knowledge across different source domains and (2) aligning label distributions between source and target domains. EDTM leverages an ensemble-based classifier expert mechanism to enhance the contribution of source domains that are more similar to the target domain. To further stabilize the learning process and improve performance, we incorporate imitation learning into the training of the target model. In addition, Maximum Classifier Discrepancy (MCD) is employed to align class-wise label distributions between the source and target domains. Experiments were conducted using Digits-Five, one of the most representative benchmark datasets for MSDA. The results show that EDTM consistently outperforms existing methods in terms of average classification accuracy. Notably, EDTM achieved significantly higher performance on target domains such as Modified National Institute of Standards and Technolog with blended background images(MNIST-M) and Street View House Numbers(SVHN) datasets, demonstrating enhanced generalization compared to baseline approaches. Furthermore, an ablation study analyzing the contribution of each loss component validated the effectiveness of the framework, highlighting the importance of each module in achieving optimal performance.
Read moreMulti-Source Domain Adaptation Using Ambient Sensor Data
Smart buildings have gained increasing interest recently by providing several advanced solutions, especially AI-based solutions. Activity recognition and occupancy estimation are among the outcomes of smart buildings that can help provide several advantages such as energy management and security solutions. Previously, domain adaptation (DA) has been widely considered by researchers to transfer knowledge from source domains, where we have abundant labeled data, to a target domain where labeled data is scarce. It is a tedious and time-consuming task to label data, especially with smart building applications which is why researchers have considered unsupervised DA where we do have labeled data in the source domain and unlabeled data in the target domain. Semi-supervised DA (SSDA) adaptation has also been considered by researchers where we have a small amount of labeled data in the target domain. Most unsupervised DA (UDA) and SSDA methods transfer knowledge from one source to one target. However, it is possible to exploit knowledge from multiple source domains instead of one single domain to enhance the performance of the target domain. Multi-source DA (MSDA) is more difficult than single-source DA but also it is more efficient. In this research, we adapt several MDSA methods and evaluate them using sensorial datasets.
Read morePrototype-Based Multisource Domain Adaptation.
Unsupervised domain adaptation aims to transfer knowledge from labeled source domain to unlabeled target domain. Recently, multisource domain adaptation (MDA) has begun to attract attention. Its performance should go beyond simply mixing all source domains together for knowledge transfer. In this article, we propose a novel prototype-based method for MDA. Specifically, for solving the problem that the target domain has no label, we use the prototype to transfer the semantic category information from source domains to target domain. First, a feature extraction network is applied to both source and target domains to obtain the extracted features from which the domain-invariant features and domain-specific features will be disentangled. Then, based on these two kinds of features, the named inherent class prototypes and domain prototypes are estimated, respectively. Then a prototype mapping to the extracted feature space is learned in the feature reconstruction process. Thus, the class prototypes for all source and target domains can be constructed in the extracted feature space based on the previous domain prototypes and inherent class prototypes. By forcing the extracted features are close to the corresponding class prototypes for all domains, the feature extraction network is progressively adjusted. In the end, the inherent class prototypes are used as a classifier in the target domain. Our contribution is that through the inherent class prototypes and domain prototypes, the semantic category information from source domains is transformed into the target domain by constructing the corresponding class prototypes. In our method, all source and target domains are aligned twice at the feature level for better domain-invariant features and more closer features to the class prototypes, respectively. Several experiments on public data sets also prove the effectiveness of our method.
Read moreDynamic Classifier Alignment for Unsupervised Multi-Source Domain Adaptation
Unsupervised domain adaptation leverages the previously gained knowledge from a labeled source domain to tackle the task from a different but similar unlabeled target domain. Most existing methods focus on transferring knowledge from a single source domain, but the information from a single domain may be inadequate to complete the target task. Some previous studies have turned to multi-view representations to enrich the transferable information. However, they simply concatenate multi-view features, which may result in information redundancy. In this paper, we propose a dynamic classifier alignment (DCA) method for multi-source domain adaptation, which aligns classifiers driven from multi-view features via a sample-wise automatic way. As proposed, both the importance of each view and the contribution of each source domain are investigated. To determine the important degrees of multiple views, an importance learning function is built by generating an auxiliary classifier. To learn the source combination parameters, a domain discriminator is developed to estimate the probability of a sample belonging to multiple source domains. Meanwhile, a self-training strategy is proposed to enhance the cross-domain ability of source classifiers with the assistance of pseudo target labels. Experiments on real-world visual datasets show the superiority of the proposed DCA.
Read moreImproving Unsupervised Domain Adaptation via Multiple Adversarial Learning
<p>Most machine learning methods assume the training and test sets to be independent and have identical distributions. However, this assumption does not always hold true in practical applications. Direct training usually induces poor performance if the training and test data have distribution shifts. To address this issue, a three-part model based on using a feature extractor, a classifier, and several domain discriminators is adopted herein. This unsupervised domain adaptation model is based on multiple adversarial learning with samples of different importance. A deep neural network is used for supervised classification learning of the source domain. Numerous adversarial networks are used to constitute the domain discriminators to align each category in the source and target domains and effectively transfer knowledge from the source domain to the target domain. Triplet loss functions&mdash;classification loss, label credibility loss, and discrimination loss&mdash;are presented to further optimize the model parameters. First, the label similarity metric is designed between the target and source domains data. Second, a credibility loss function is proposed to obtain an accurate label for the unlabeled data of the target domain under training iterations. Finally, a discrimination loss is designed for multiple adversarial domain discriminators to fully utilize the unlabeled data in the target domain during training. The discrimination loss function uses predicted label probabilities as dynamic weights for the train data. The proposed method is compared with mainstream domain adaptation approaches on four public datasets: Office-31, MNIST, USPS, and SVHN. Experimental results show that the proposed method can perform well in the target domain and improve generalization performance of the model.</p> <p>&nbsp;</p>
Read moreTight Source Domain Match for Partial Domain Adaptation Based on Maximum Density
Unsupervised domain adaptation (UDA) aims at transferring knowledge between a well-labelled ’source domain’ and an unlabelled ’target domain’ by decreasing distribution discrepancy. In real scenario, partial domain adaptation (PDA), where target domain only includes part of the classes of source domain, is adopted as fully-shared label space is often unavailable. Non-identical label spaces across domains lead to performance degradation due to source-unique classes being mis-matched to the target domain, i.e. negative transfer of the target domain. Although existing PDA approaches have produced promising results, they still confront with negative transfer problem without rigorous generalization bounds. In our work, a novel PDA model has been proposed based on margin disparity and maximum source intra-class density divergence (MDSD). It matches the feature distributions with shared labels and congregates source samples in the source with affirmative labels. Removing maximum target density with pseudo labels, it effectively avoids over-fitting and accelerates learning speed. In addition, we construct a new garbage classification dataset, which is comprised of source domain - Product (Pr) and target domain - Garbage (Ga) for PDA validation. Experiments show that our proposed unsupervised partial domain adaptation method has a good performance on Office-31, Office-Home and Pr-Ga datasets.
Read moreMulti-Source Unsupervised Domain Adaptation via Pseudo Target Domain.
Multi-source domain adaptation (MDA) aims to transfer knowledge from multiple source domains to an unlabeled target domain. MDA is a challenging task due to the severe domain shift, which not only exists between target and source but also exists among diverse sources. Prior studies on MDA either estimate a mixed distribution of source domains or combine multiple single-source models, but few of them delve into the relevant information among diverse source domains. For this reason, we propose a novel MDA approach, termed Pseudo Target for MDA (PTMDA). Specifically, PTMDA maps each group of source and target domains into a group-specific subspace using adversarial learning with a metric constraint, and constructs a series of pseudo target domains correspondingly. Then we align the remainder source domains with the pseudo target domain in the subspace efficiently, which allows to exploit additional structured source information through the training on pseudo target domain and improves the performance on the real target domain. Besides, to improve the transferability of deep neural networks (DNNs), we replace the traditional batch normalization layer with an effective matching normalization layer, which enforces alignments in latent layers of DNNs and thus gains further promotion. We give theoretical analysis showing that PTMDA as a whole can reduce the target error bound and leads to a better approximation of the target risk in MDA settings. Extensive experiments demonstrate PTMDA's effectiveness on MDA tasks, as it outperforms state-of-the-art methods in most experimental settings.
Read moreEnhancing Cross-Domain Remote Sensing Scene Classification by Multi-Source Subdomain Distribution Alignment Network
Multi-source domain adaptation (MSDA) in remote sensing (RS) scene classification has recently gained significant attention in the visual recognition community. It leverages multiple well-labeled source domains to train a model capable of achieving strong generalization on the target domain with little to no labeled data from the target domain. However, the distribution shifts among multiple source domains make it more challenging to align the distributions between the target domain and all source domains concurrently. Moreover, relying solely on global alignment risks losing fine-grained information for each class, especially in the task of RS scene classification. To alleviate these issues, we present a Multi-Source Subdomain Distribution Alignment Network (MSSDANet), which introduces novel network structures and loss functions for subdomain-oriented MSDA. By adopting a two-level feature extraction strategy, this model attains better global alignment between the target domain and multiple source domains, as well as alignment at the subdomain level. First, it includes a pre-trained convolutional neural network (CNN) as a common feature extractor to fully exploit the shared invariant features across one target and multiple source domains. Secondly, a dual-domain feature extractor is used after the common feature extractor, which maps the data from each pair of target and source domains to a specific dual-domain feature space and performs subdomain alignment. Finally, a dual-domain feature classifier is employed to make predictions by averaging the outputs from multiple classifiers. Accompanied by the above network, two novel loss functions are proposed to boost the classification performance. Discriminant Semantic Transfer (DST) loss is exploited to force the model to effectively extract semantic information among target and source domain samples, while Class Correlation (CC) loss is introduced to reduce the feature confusion from different classes within the target domain. It is noteworthy that our MSSDANet is developed in an unsupervised manner for domain adaptation, indicating that no label information from the target domain is required during training. Extensive experiments on four common RS image datasets demonstrate that the proposed method achieves state-of-the-art performance for cross-domain RS scene classification. Specifically, in the dual-source and three-source settings, MSSDANet outperforms the second-best algorithm in terms of overall accuracy (OA) by 2.2% and 1.6%, respectively.
Read moreMulti-Source Domain Adaptation with Collaborative Learning for Semantic Segmentation
Multi-source unsupervised domain adaptation (MSDA) aims at adapting models trained on multiple labeled source domains to an unlabeled target domain. In this paper, we propose a novel multi-source domain adaptation framework based on collaborative learning for semantic segmentation. Firstly, a simple image translation method is introduced to align the pixel value distribution to reduce the gap between source domains and target domain to some extent. Then, to fully exploit the essential semantic information across source domains, we propose a collaborative learning method for domain adaptation without seeing any data from target domain. In addition, similar to the setting of unsupervised domain adaptation, unlabeled target domain data is leveraged to further improve the performance of domain adaptation. This is achieved by additionally constraining the outputs of multiple adaptation models with pseudo labels online generated by an ensembled model. Extensive experiments and ablation studies are conducted on the widely-used domain adaptation benchmark datasets in semantic segmentation. Our proposed method achieves 59.0% mIoU on the validation set of Cityscapes by training on the labeled Synscapes and GTA5 datasets and unlabeled training set of Cityscapes. It significantly outperforms all previous state-of-the-arts single-source and multi-source unsupervised domain adaptation methods.
Read morePartial Feature Selection and Alignment for Multi-Source Domain Adaptation
Multi-Source Domain Adaptation (MSDA), which dedicates to transfer the knowledge learned from multiple source domains to an unlabeled target domain, has drawn increasing attention in the research community. By assuming that the source and target domains share consistent key feature representations and identical label space, existing studies on MSDA typically utilize the entire union set of features from both the source and target domains to obtain the feature map and align the map for each category and domain. However, the default setting of MSDA may neglect the issue of "partialness", i.e., 1) a part of the features contained in the union set of multiple source domains may not present in the target domain; 2) the label space of the target domain may not completely overlap with the multiple source domains. In this paper, we unify the above two cases to a more generalized MSDA task as Multi-Source Partial Domain Adaptation (MSPDA). We propose a novel model termed Partial Feature Selection and Alignment (PFSA) to jointly cope with both MSDA and MSPDA tasks. Specifically, we firstly employ a feature selection vector based on the correlation among the features of multiple sources and target domains. We then design three effective feature alignment losses to jointly align the selected features by preserving the domain information of the data sample clusters in the same category and the discrimination between different classes. Extensive experiments on various benchmark datasets for both MSDA and MSPDA tasks demonstrate that our proposed PFSA approach remarkably outperforms the state-of-the-art MSDA and unimodal PDA methods.
Read moreModeling Sequential Domain Shift through Estimation of Optimal Sub-spaces for Categorization
Domain adaptation (DA) is the process in which labeled training samples available from one domain is used to improve the performance of statistical tasks performed on test samples drawn from a different domain. The domain from which the training samples are obtained is termed as the source domain, and the counterpart consisting of the test samples is termed as the target domain. Few unlabeled training samples are also taken from the target domain in order to approximate its distribution. In this paper, we propose a new method of unsupervised DA, where a set of domain invariant sub-spaces are estimated using the geometrical and statistical properties of the source and target domains. This is a modification of the work done by Gopalan et al. [2], where the geodesic path from the principal components of the source to that of the target is considered in the Grassmann manifold, and the intermediary points are sampled to represent the incremental change in the geometric properties of the data in source and target domains. Instead of the geodesic path, we consider an alternate path of shortest length between the principal components of source and target, with the property that the intermediary sample points on the path form domain invariant sub-spaces using the concept of Maximum Mean Discrepancy (MMD) [3]. Thus we model the change in the geometric properties of data in both the domains sequentially, in a manner such that the distributions of projected data from both the domains always remain similar along the path. The entire formulation is done in the kernel space which makes it more robust to non-linear transformations. Let X and Y be the source and target domains having nX and nY number of instances respectively. If Φ(.) is a universal kernel function, then in kernel space the source and target domains are Φ(X) ∈ RnX×d and Φ(Y ) ∈ RnX×d respectively. Let KXX and KYY be the kernel gram matrices of Φ(X) and Φ(Y ) respectively. Let D = [X ;Y ] denote the combined source and target domain data, and the corresponding data in kernel space is given as Φ(D). The kernel gram matrix formed using D is given by
Read more