Target Domain Features Research Articles

Unsupervised domain adaptation via softmax-based prototype construction and adaptation

To mitigate the distribution difference between the source and target domains, there have been many unsupervised domain adaptation methods to achieve class-level alignment by aligning the prototypes of two domains. Since the labels of the target domain are unobserved, the target prototypes are constructed based on pseudo-labels. However, the inaccuracy of pseudo-labels may lead to biased prototypes, which introduce noise into the distribution alignment. Moreover, a shared feature extractor or two separate feature extractors are used to extract domain-invariant features. But the former is limited by the large domain gap and the latter increases the network parameters. To this end, we propose a Softmax-Based Prototype construction and Adaptation (SBPA) method, which constructs prototypes based on the softmax output of the classifier instead of ground-truth labels or pseudo-labels. SBPA performs domain-level alignment through adversarial training, and class-level alignment by aligning prototypes of the same class. In addition, SBPA contains a residual block to explicitly model the difference between the source and target domain features extracted by a shared feature extractor. We evaluate our method on four widely used datasets, and the results show that our method outperforms recent domain adaptation methods, especially on DomainNet, the hardest domain adaptation dataset by far.

Clustering-Guided Novel Unsupervised Domain Adversarial Network for Partial Transfer Fault Diagnosis of Rotating Machinery

Unsupervised partial transfer fault diagnosis studies of rotating machinery have practical significance, which still exists some challenges, for example, the learned domain-specific statistics and parameters usually influence the learning effect of target-domain features to some degree, and the relatively scattered target-domain features will lead to negative transfer. To overcome those limitations and further improve partial transfer fault diagnosis performance, a clustering-guided novel unsupervised domain adversarial network is proposed in this paper. Firstly, a novel unsupervised domain adversarial network is constructed using domain-specific batch normalization to remove domain-specific information to enhance alignment between source and target domains. Secondly, embedded clustering strategy is designed to learn tightly clustered target-domain features to suppress negative transfer in partial domain adaptation process. Finally, a joint optimization objective function is defined to balance different losses to improve the training and diagnosis performance. Two experimental cases of bevel gearbox and bearing are used to validate the effectiveness and superiority of the proposed method in solving unsupervised partial transfer fault diagnosis problems.

ADAN: An Adversarial Domain Adaptation Neural Network for Early Gastric Cancer Prediction.

Gastric cancer is a highly prevalent cancer world-wide. Accurate diagnosis of Early Gastric Cancer (EGC) is of great significance to improve the treatment and survival rate of patients. However, EGC and gastric ulcers have similar en-doscopic image characteristics, resulting in a high misdiagnosis rate. Most existing deep learning and machine learning models for EGC recognition have the disadvantages of cumbersome pre-processing steps and high leakage ratios. To address the above challenges, we propose an end-to-end Adversarial Do-main Adaptation Neural network (ADAN) for EGC prediction on endoscopic images. ADAN network consists of a source domain feature extractor, a source domain classifier, two target domain feature extractors, a target domain classifier, and a domain discriminator. A source domain feature extractor is designed to train the model on public gastrointestinal datasets, which effectively solves the problem of insufficient training data. In addition, an adaptive source-target domain mapping classifier is added to each target domain feature extractor for automatically adjusting the number of classification categories in the target domain. Experimental results show that the proposed ADAN network is superior to the most advanced methods and can accurately predict EGC in clinical practice. Clinical relevance-In this study, the EGC diagnosis model based on the adversarial domain adaptive construction will be more applicable to the real clinical scenario, with higher accuracy and sensitivity and assist the endoscopist to make more accurate diagnosis for EGC and reduce the workload.

Cross-Domain Remaining Useful Life Prediction Based on Adversarial Training

Remaining useful life prediction can assess the time to failure of degradation systems. Currently, numerous neural network-based prediction methods have been proposed by researchers. However, most of the work contains an implicit prerequisite: the network training and testing data have the same operating conditions. To solve this problem, an adversarial discriminative domain adaption prediction method based on adversarial training is proposed to improve the accuracy of cross-domain prediction under different working conditions. First, an LSTM feature extraction network is constructed to mine the source domain data and the target domain data for deep feature representation. Subsequently, the parameters of the target domain feature extraction network are adjusted based on the idea of adversarial training to achieve domain invariant feature mining. The proposed scheme is experimented on a publicly available dataset and achieves state-of-the-art prediction performance compared to recent unsupervised domain adaptation prediction methods.

Credit Risk Modeling Using Transfer Learning and Domain Adaptation.

In the domain of credit risk assessment lenders may have limited or no data on the historical lending outcomes of credit applicants. Typically this disproportionately affects Micro, Small, and Medium Enterprises (MSMEs), for which credit may be restricted or too costly, due to the difficulty of predicting the Probability of Default (PD). However, if data from other related credit risk domains is available Transfer Learning may be applied to successfully train models, e.g., from the credit card lending and debt consolidation (CD) domains to predict in the small business lending domain. In this article, we report successful results from an approach using transfer learning to predict the probability of default based on the novel concept of Progressive Shift Contribution (PSC) from source to target domain. Toward real-world application by lenders of this approach, we further address two key questions. The first is to explain transfer learning models, and the second is to adjust features when the source and target domains differ. To address the first question, we apply Shapley values to investigate how and why transfer learning improves model accuracy, and also propose and test a domain adaptation approach to address the second. These results show that adaptation improves model accuracy in addition to the improvement from transfer learning. We extend this by proposing and testing a combined strategy of feature selection and adaptation to convert values of source domain features to better approximate values of target domain features. Our approach includes a strategy to choose features for adaptation and an algorithm to adapt the values of these features. In this setting, transfer learning appears to improve model accuracy by increasing the contribution of less predictive features. Although the percentage improvements are small, such improvements in real world lending could be of significant economic importance.

Cross-domain Mutual Information Adversarial Maximization

Domain adaptation challenges the problem where the source domain and the target domain have distinctive data distributions. Different from previous approaches which align the two domains by minimizing a distribution metric, in this paper, we report a new perspective of handling unsupervised domain adaptation. Specifically, we formulate domain adaptation as maximizing the obtained knowledge of the target domain through observing the source domain. Technically, we maximize the mutual information between the source domain features and the target domain features in a deep adversarial network. Firstly, we use a feature extraction network and a domain discriminator with opposite goals to form adversarial components, and learn the domain-invariant features between the source and target domains through adversarial training. Secondly, we use the optimization goal of maximizing the mutual information between cross-domain features to supervise the adversarial training process to ensure that the maximum target domain information can be obtained by observing the source domain features. Finally, we evaluate our method on four datasets: Office-31, ImageCLEF-DA, Office-Home, and VisDA-2017, and all achieve better performance than previous methods. We show that our method, named Cross-domain Mutual Information Adversarial Maximization (CMIAM), is a promising approach and able to outperform previous state-of-the-arts on various unsupervised domain adaptation tasks.

Domain Adaptation via a Task-Specific Classifier Framework for Remote Sensing Cross-Scene Classification

The scene classification of high spatial resolution (HSR) imagery involves labeling an HSR image with a specific high-level semantic class according to the composition of the semantic objects and their spatial relationships. As such, scene classification has attracted increased attention in recent years, and many different algorithms have now been proposed for the cross-scene classification task. However, the recently proposed scene classification methods based on deep convolutional neural networks (CNNs) still suffer from domain shift problems, because of the training data and validation data not following the assumption of independent and identical distributions. The employment of generative adversarial networks has been found to be an effective way to bridge the domain shift/gap. However, the existing cross-scene classification methods do not use the classification information in the target domain, and the domain classifier is task-independent for different scene classification tasks. In this article, to solve this problem, domain adaptation via a task-specific classifier (DATSNET) framework is proposed for HSR image scene classification. Task-specific classifiers and minimizing and maximizing, “ i.e., minimaxing,” of the classifier discrepancy are integrated in the DATSNET framework. The task-specific classifiers are proposed to align the distributions of the source domain features and target domain features by utilizing task-specific decision boundaries in the target domain. In order to align the two task-specific classifiers’ feature distributions, minimaxing the defined discrepancy between the different classifiers in an adversarial manner is proposed to obtain better task-specific classifier boundaries in the target domain and a better-aligned feature distribution in both domains. The experimental results obtained with different remote sensing cross-scene classification tasks demonstrate that the proposed method achieves a significantly improved performance compared with the other state-of-the-art remote sensing cross-scene classification algorithms.

Stagewise Unsupervised Domain Adaptation With Adversarial Self-Training for Road Segmentation of Remote-Sensing Images

Road segmentation from remote sensing images is a challenging task with wide ranges of application potentials. Deep neural networks have advanced this field by leveraging the power of large-scale labeled data, which, however, are extremely expensive and time-consuming to acquire. One solution is to use cheap available data to train a model and deploy it to directly process the data from a specific application domain. Nevertheless, the well-known domain shift (DS) issue prevents the trained model from generalizing well on the target domain. In this paper, we propose a novel stagewise domain adaptation model called RoadDA to address the DS issue in this field. In the first stage, RoadDA adapts the target domain features to align with the source ones via generative adversarial networks (GAN) based inter-domain adaptation. Specifically, a feature pyramid fusion module is devised to avoid information loss of long and thin roads and learn discriminative and robust features. Besides, to address the intra-domain discrepancy in the target domain, in the second stage, we propose an adversarial self-training method. We generate the pseudo labels of target domain using the trained generator and divide it to labeled easy split and unlabeled hard split based on the road confidence scores. The features of hard split are adapted to align with the easy ones using adversarial learning and the intra-domain adaptation process is repeated to progressively improve the segmentation performance. Experiment results on two benchmarks demonstrate that RoadDA can efficiently reduce the domain gap and outperforms state-of-the-art methods.

Adversarial Learning Based Discriminative Domain Adaptation for Geospatial Image Analysis

The ability of supervised image analysis methods to provide state-of-the-art performance is limited by availability of high-quality labeled data in large quantities. Domain adaptation approaches propose a solution to this problem by leveraging quality labeled information from auxiliary data sources. In this work, we use adversarial learning for domain adaptation for remote sensing applications. First, we approached the problem of unavailable target domain labels with unsupervised domain adaptation and then extended our method for semisupervised domain adaptation to use a few available labels as well. We are using adversarial learning to extract discriminative target domain features that are aligned with source domain. We test our framework for two very different applications of remote sensing imagery, multiclass classification in hyperspectral images and semantic segmentation in large scale satellite images. For hyperspectral image analysis two datasets were used: the University of Houston shadow data was used for quantifying the efficacy of our approach to varying illumination, and the Botswana data was used to quantify the efficacy of our approach under multitemporal spectral shifts. Multisensor high-resolution images from National Agriculture Imagery Program and SpaceNet-Rio datasets were used as the source and target for the task of building extraction for large scale semantic segmentation based domain adaptation.

RFA-Net: Reconstructed Feature Alignment Network for Domain Adaptation Object Detection in Remote Sensing Imagery

With the development of deep learning, great progress has been made in object detection of remote sensing (RS) imagery. However, the object detector is hard to generalize well from one labeled dataset (source domain) to another unlabeled dataset (target domain) due to the discrepancy of data distribution (domain shift). Currently, adversarial-based domain adaptation methods align the semantic features of source and target domain features to alleviate the domain shift. But they fail to avoid the alignment of noisy background features and neglect the instance-level features, which are inappropriate for detection models that focus on instance location and classification. To mitigate domain shift existing in object detection, we propose a reconstructed feature alignment network (RFA-Net) for unsupervised cross-domain object detection in remote sensing imagery. The RFA-Net includes one sequential data augmentation module (SDA) deployed on data level for providing solid gains on unlabeled data, one sparse feature reconstruction module (SFR) deployed on feature level to intensify instance feature for feature alignment, and one pseudo-label generation module (PLG) deployed on label level for the supervision of the unlabeled target domain. Extensive experiments illustrate that our proposed RFA-Net is effective to alleviate the domain shift problem in domain adaptation object detection of RS imagery.

Synthetic Source Universal Domain Adaptation through Contrastive Learning.

Universal domain adaptation (UDA) is a crucial research topic for efficient deep learning model training using data from various imaging sensors. However, its development is affected by unlabeled target data. Moreover, the nonexistence of prior knowledge of the source and target domain makes it more challenging for UDA to train models. I hypothesize that the degradation of trained models in the target domain is caused by the lack of direct training loss to improve the discriminative power of the target domain data. As a result, the target data adapted to the source representations is biased toward the source domain. I found that the degradation was more pronounced when I used synthetic data for the source domain and real data for the target domain. In this paper, I propose a UDA method with target domain contrastive learning. The proposed method enables models to leverage synthetic data for the source domain and train the discriminativeness of target features in an unsupervised manner. In addition, the target domain feature extraction network is shared with the source domain classification task, preventing unnecessary computational growth. Extensive experimental results on VisDa-2017 and MNIST to SVHN demonstrated that the proposed method significantly outperforms the baseline by 2.7% and 5.1%, respectively.

The manifold embedded selective pseudo-labeling algorithm and transfer learning of small sample dataset

Special scene classification and identification tasks are not easily fulfilled to obtain samples, which results in a shortage of samples. The focus of current researches lies in how to use source domain data (or auxiliary domain data) to build domain adaption transfer learning models and to improve the classification accuracy and performance of small sample machine learning in these special and difficult scenes. In this paper, a model of deep convolution and Grassmann manifold embedded selective pseudo-labeling algorithm (DC-GMESPL) is proposed to enable transfer learning classifications among multiple small sample datasets. Firstly, DC-GMESPL algorithm uses satellite remote sensing image sample data as the source domain to extract the smoke features simultaneously from both the source domain and the target domain based on the Resnet50 deep transfer network. This is done for such special scene of the target domain as the lack of local sample data for forest fire smoke video images. Secondly, DC-GMESPL algorithm makes the source domain feature distribution aligned with the target domain feature distribution. The distance between the source domain and the target domain feature distribution is minimized by removing the correlation between the source domain features and re-correlation with the target domain. And then the target domain data is pseudo-labeled by selective pseudo-labeling algorithm in Grassmann manifold space. Finally, a trainable model is constructed to complete the transfer classification between small sample datasets. The model of this paper is evaluated by transfer learning between satellite remote sensing image and video image datasets. Experiments show that DC-GMESPL transfer accuracy is higher than DC-CMEDA, Easy TL, CMMS and SPL respectively. Compared with our former DC-CMEDA, the transfer accuracy of our new DC-GMESPL algorithm has been further improved. The transfer accuracy of DC-GMESPL from satellite remote sensing image to video image has been improved by 0.50%, the transfer accuracy from video image to satellite remote sensing image has been improved by 8.50% and then, the performance has been greatly improved.

A Transfer Learning Based Unmanned Aerial Vehicle MEMS Inertial Sensors Fault Diagnosis Method

In this paper, we propose a novel transfer learning based micro-electromechanical system (MEMS) inertial sensors fault diagnosis method. First, the MEMS inertial sensors fault diagnosis method is formulated to a deep transfer learning problem in which the offline samples are deemed as source domain and the online samples are set to target domain features. Second, the bidirectional long short-term memory and Hilbert-Huang transformation-based feature transfer model is designed to decrease the discrepancy between SD and TD, that performs the transfer operation using intrinsic mode function features. Then we propose a convolutional neuro network-based transfer learning algorithm to further decrease deep features discrepancy and perform the fault classification tasks on TD. According to the experiments, the proposed FD method has achieved excellent fault classification performance and significantly improvement comparing with the state-of-the art methods.

Model Parameter Transfer for Gear Fault Diagnosis under Varying Working Conditions

Gear fault diagnosis technologies have received rapid development and been effectively implemented in many engineering applications. However, the various working conditions would degrade the diagnostic performance and make gear fault diagnosis (GFD) more and more challenging. In this paper, a novel model parameter transfer (NMPT) is proposed to boost the performance of GFD under varying working conditions. Based on the previous transfer strategy that controls empirical risk of source domain, this method further integrates the superiorities of multi-task learning with the idea of transfer learning (TL) to acquire transferable knowledge by minimizing the discrepancies of separating hyperplanes between one specific working condition (target domain) and another (source domain), and then transferring both commonality and specialty parameters over tasks to make use of source domain samples to assist target GFD task when sufficient labeled samples from target domain are unavailable. For NMPT implementation, insufficient target domain features and abundant source domain features with supervised information are fed into NMPT model to train a robust classifier for target GFD task. Related experiments prove that NMPT is expected to be a valuable technology to boost practical GFD performance under various working conditions. The proposed methods provides a transfer learning-based framework to handle the problem of insufficient training samples in target task caused by variable operation conditions.

A CCA-Based Item-Side Alignment Method for Cross-Domain Recommendation System

For cross-domain recommendation, it can be divided into strong correlation and weak correlation problems according to the consistency between auxiliary domain and target domain. The weak correlation problem is more practical than the strong correlation problem, and the solution is more difficult. The difficulty lies in how to establish an effective transfer model, to make sure the auxiliary domain and the target domain can perform effective collaborative training. For weak correlation problem, if the item-side of auxiliary domain and the target domain are not aligned, or the transfer model has a strong dependency on the user-side of the auxiliary domain, it will seriously affect the effect of cross-domain recommendation. To solve these problems mentioned above, we propose a CCA-based item-side alignment method (CIAM) by introducing: (1) item side alignment method. We use CCA to align the item side between auxiliary and target domain, to intensify the weak correlation between 2 domains. (2) the transfer model of retaining the user feature of target domain. The CIAM retained user features of target domain in UV decomposition, that makes the transfer model could not destroy the user feature between 2 domains. The proposed CIAM can improve the assistance of auxiliary domain, and can avoid the influence of the needless user-side of the auxiliary do-main on cross-domain recommendation. By experimental analysis, it can be verified that the proposed CIAM algorithm has a better performance than general cross-domain recommendation methods.

Interpretable Rule-Based Fuzzy ELM and Domain Adaptation for Remote Sensing Image Classification

The size of the remote sensing (RS) data is increasing exponentially with modern technological evolution. Efficient mapping with these data using a supervised model demands ample labeled samples, and obtaining them is expensive and time-consuming. For that reason, domain adaptation (DA) methodologies that work with limited available knowledge in terms of labeled samples or trained models are highly successful. DA uses the model trained for a domain called “source” and aims to classify the data of a related domain called “target.” However, an insufficient amount of labeled samples in the target domain data set is still a bottleneck for DA models used for RS images. To address this, we have proposed an efficient DA classification model using an interpretable rule-based fuzzy extreme learning machine (IRF-ELM). These rules are derived using the maximum fuzzy membership value of features characterized by class-belonging fuzzification and two rule extraction matrices. A partially connected ELM architecture is then designed with these rules. The fuzzification process typically aligns the target domain's feature space to the source domain in the best possible ways to transfer knowledge between them. The proposed model's superiority to similar other methods is verified using different RS data sets in terms of various performance measurement indexes.

Transfer Learning Based Method for Frequency Response Model Updating with Insufficient Data.

Finite element model updating precision depends heavily on sufficient vibration feature extraction. However, adequate amount of sample collection is generally time-consuming in frequency response (FR) model updating. Accurate vibration feature extraction with insufficient data has become a significant challenge in FR model updating. To update the finite element model with a small dataset, a novel approach based on transfer learning is firstly proposed in this paper. A readily available fault diagnosis dataset is selected as ancillary knowledge to train a high-precision mapping from FR data to updating parameters. The proposed transfer learning network is constructed with two branches: source and target domain feature extractor. Considering about the cross-domain feature discrepancy, a domain adaptation method is designed by embedding the extracted features into a shared feature space to train a reliable model updating framework. The proposed method is verified by a simulated satellite example. The comparison results manifest that sample amount dependency has prominently lessened this method and the updated model outperforms the method without transfer learning in accuracy with the small dataset. Furthermore, the updated model is validated through dynamic response out of the training set.

Domain Adaptation-Based Transfer Learning for Gear Fault Diagnosis Under Varying Working Conditions

The nonidentical and nonindependent distribution problem, caused by various working conditions of gears, inevitably makes the effectiveness of gear fault diagnosis (GFD) degrade because of insufficient or low-quality data. Although some domain adaptation learning (DAL) models exist, most of them only aim at minimizing the global distributional mean discrepancy between the source domain (SD) and the target domain (TD), while the contribution of individual data is neglected. To address this issue, a new DAL, aiming to reduce the discrepancy on extracted features under a least square support vector machine (LSSVM) framework, is studied to exploit SD signals from another working conditions or adjacent mechanical parts to assist and boost target gear fault diagnostic performance in this article. In addition, the white cosine similarity criterion is adopted to quantize the distributional weights of SD and TD feature data, and then the weights are added into a regularization function that measures the projected distributional discrepancy in the LSSVM framework. Related experiments prove that the proposed method has higher diagnostic accuracies than other models. So, this strategy is expected to be a useful tool to transfer from SD to TD, and to boost the GFD performance under various working conditions.

Hyperspectral Image Classification Based on Unsupervised Heterogeneous Domain Adaptation CycleGan

Aiming at the difficulty of obtaining sufficient labeled Hyperspectral image (HSI) data and the inconsistent feature distribution of different HSIs, a novel Unsupervised heterogeneous domain adaptation CycleGan (UHDAC) is proposed by using CycleGan to capture the transferable features in the absence of similar data. On the one hand, the two-way mapping is used to find the internal relationship between the source and target domain data, while the two-way adversary is used to constrain the source and target domain features, realizing the alignment of feature distributions. On the other hand, the CORAL loss function is introduced to minimize the distance between the second-order statistical difference between the source and target domain features, so as to solve the insufficient constraint of mapping relationship caused by the low consistency of HSI data structure in different domains. Experiments on three real HSI datasets show that UHDAC can effectively realize the unsupervised classification of target domain HSI with high classification accuracy by using the labeled HSI data in the source domain.

Unsupervised image-to-image translation using intra-domain reconstruction loss

Generative adversarial networks (GANs) have been successfully used for considerable computer vision tasks, especially the image-to-image translation. However, GANs are often accompanied by training instability and mode collapse in the process of image-to-image translation, which leads to the generation of low-quality images. To address the aforementioned problem, by combining CycleGAN and intra-domain reconstruction loss (IDRL), we propose an unsupervised image-to-image translation network named “Cycle-IDRL”. Specifically, the generator adopts the U-Net network with skip connections, which merges the coarse-grained and fine-grained features and the least squares loss in LSGAN is used to improve the stability of training process. Especially, the target domain features extracted from the discriminator are used as input of generator to generate reconstructed samples. Then, we construct the IDRL between the target domain samples and the reconstructed samples by using L1 norm. The experimental results on multiple datasets show that the proposed method performs better than the existing methods.