Domain Adaptation Research Articles

Nonconvex and discriminative transfer subspace learning for unsupervised domain adaptation

Nonconvex and discriminative transfer subspace learning for unsupervised domain adaptation

Study on Image Classification Algorithm Based on Multi-Scale Feature Fusion and Domain Adaptation

This paper introduces the MMTADAN, an innovative algorithm designed to enhance cross-domain image classification. By integrating multi-scale feature extraction with Taylor series-based detail enhancement and adversarial domain adaptation, the MMTADAN effectively aligns features between the source and target domains. The proposed approach addresses the critical challenge of generalizing classification models across diverse datasets, demonstrating significant improvements in performance. The findings suggest that retaining essential image details through multi-scale extraction and Taylor series enhancement can lead to better classification outcomes, making the MMTADAN a valuable contribution to the field of image classification.

Fault diagnosis of rolling bearings driven by multi-channel data fusion and feature fusion under time-varying speed conditions

Abstract Bearings, as the core component for power transmission, are crucial in ensuring the safe and reliable operation of equipment. However, the fault information contained in a single-channel vibration signal is inherently limited. Additionally, under time-varying speed conditions, features are prone to drift, and the cross-domain diagnostic performance of most traditional domain adaptation (DA) models may drop dramatically. To solve the above problems and enhance the ability of DA models in extracting domain invariant features, this paper introduces a Multi-channel data fusion and Attention-guided Multi-feature Fusion-driven Center-aligned Network (MAMC). Initially, a multi-channel time-frequency information fusion strategy based on wavelet transform is constructed to achieve a comprehensive fusion of multi-channel data, thereby obtaining richer fault feature representations. Subsequently, a multi-branch feature fusion network, integrated with an attention mechanism, is devised to capture significant features across various dimensions and scales, resulting in more comprehensive and representative fault features. Finally, a novel Center-Aligned Domain Adaptation method (CADA) is proposed based on domain adversarial methods and center loss. By minimizing the distance between deep domain invariant features and trainable common class centers, the issue of domain shift between data is effectively alleviated, and the cross-domain diagnostic performance of DA models under the time-varying speed conditions is improved. The experimental results indicate that the MAMC method exhibits superior performance on both bearing datasets and is a promising approach for cross-domain intelligent fault diagnosis.

Target-aware cross-modality unsupervised domain adaptation for vestibular schwannoma and cochlea segmentation.

There is growing interest in research on segmentation for the vestibular schwannoma (VS) and cochlea using high-resolution T2 (hrT2) imaging over contrast-enhanced T1 (ceT1) imaging due to the contrast agent side effects. However, the hrT2 imaging remains a problem of insufficient annotated data, which is fatal for building more robust segmentation models. To address the issue, recent studies have adopted unsupervised domain adaptation approaches that translate ceT1 images to hrT2 images. However, previous studies did not consider the size and visual characteristics of the target objects, such as VS and cochlea, during image translation. Specifically, those works simply performed normalization on the entire image without considering its significant impact on the quality of the translated images. These approaches tend to erase the small target objects, making it difficult to preserve the structure of these objects when generating pseudo-target images. Furthermore, they may also struggle to accurately reflect the unique style of the target objects within the images. Therefore, we propose a target-aware unsupervised domain adaptation framework, designed for translating target objects, each tailored to their unique visual characteristics and size using target-aware normalization. We demonstrate the superiority of the proposed framework on a publicly available challenge dataset. Codes are available at https://github.com/Bokyeong-Kang/TANQ .

Development of a Cost-Efficient and Glaucoma-Specialized OD/OC Segmentation Model for Varying Clinical Scenarios

Most existing optic disc (OD) and cup (OC) segmentation models are biased to the dominant size and easy class (normal class), resulting in suboptimal performances on glaucoma-confirmed samples. Thus, these models are not optimal choices for assisting in tracking glaucoma progression and prognosis. Moreover, fully supervised models employing annotated glaucoma samples can achieve superior performances, although restricted by the high cost of collecting and annotating the glaucoma samples. Therefore, in this paper, we are dedicated to developing a glaucoma-specialized model by exploiting low-cost annotated normal fundus images, simultaneously adapting various common scenarios in clinical practice. We employ a contrastive learning and domain adaptation-based model by exploiting shared knowledge from normal samples. To capture glaucoma-related features, we utilize a Gram matrix to encode style information and the domain adaptation strategy to encode domain information, followed by narrowing the style and domain gaps between normal and glaucoma samples by contrastive and adversarial learning, respectively. To validate the efficacy of our proposed model, we conducted experiments utilizing two public datasets to mimic various common scenarios. The results demonstrate the superior performance of our proposed model across multi-scenarios, showcasing its proficiency in both the segmentation- and glaucoma-related metrics. In summary, our study illustrates a concerted effort to target confirmed glaucoma samples, mitigating the inherent bias issue in most existing models. Moreover, we propose an annotation-efficient strategy that exploits low-cost, normal-labeled fundus samples, mitigating the economic- and labor-related burdens by employing a fully supervised strategy. Simultaneously, our approach demonstrates its adaptability across various scenarios, highlighting its potential utility in both assisting in the monitoring of glaucoma progression and assessing glaucoma prognosis.

Target Structure Learning Framework for Unsupervised Multi-Class Domain Adaptation

Unsupervised multi-class domain adaptation (multi-class UDA) has recently been proposed to fill the gap between empirically practical methods for multi-class classification and well-founded theory with the setting of binary classification. Nevertheless, the multi-class UDA methods use model predictions to characterize the disagreement of multi-class scoring hypotheses, which is used to optimize the divergence between domain distributions. Such self-training manner may bring inaccurate model predictions, which would damage the target structure due to the absence of labels of the target domain, leading to sub-optimal performance. On the other hand, this disagreement between multi-class scoring hypotheses does not involve the relationships among all of the multiple classes. It causes that multi-class UDA cannot properly connect the advanced practical UDA methods that consider class-conditional distribution alignment. Thus, we propose to exploit the target structure information and then incorporate it into multi-class UDA to achieve class-conditional distribution alignment. We theoretically and experimentally explain the importance of accurate target structure information to reduce the expected error on the target domain. Notably, our method achieves state-of-the-art results on three commonly-used benchmarks with different scales. In addition, using the target structure information, we propose a variant to cope with noisy open-world source domains such as noisy labels and out-of-distribution samples, enhancing the robustness of our method. The source code is available at https://github.com/jingzhengli/Multi_Class_UDA .

Theory and algorithm for domain adaptation

Domain adaptation is an important subfield of transfer learning, it has been successfully applied in many applications of machine learning. Recently, significant theoretical and algorithmic advances have been achieved in domain adaptation. The theoretical analyses for domain adaptation are based on VC dimension and Rademacher complexity. There are also some covering number-based results, but most of these bounds are based on the results of Rademacher complexity, indirectly given by the relationship between covering number and Rademacher complexity. In this paper, we propose a theoretical analysis framework for domain adaptation, thus the error bound can be derived directly by covering number, which is an effective method for analyzing the generalization error in statistical learning theory. We derive generalization error bound for domain adaptation with a class of loss functions satisfying the assumptions. We also propose a mixup contrastive adversarial network for domain adaptation by introducing a mixup module for enhancing the alignment of the source and target domains during domain transfer, and a contrastive learning module for solving class-level alignment after domain transfer. Experimental results demonstrate the effectiveness of the proposed algorithm and the property of the theoretical results.

A class alignment network based on self-attention for cross-subject EEG classification.

Due to the inherent variability in EEG signals across different individuals, domain adaptation and adversarial learning strategies are being progressively utilized to develop subject-specific classification models by leveraging data from other subjects. These approaches primarily focus on domain alignment and tend to overlook the critical task-specific class boundaries. This oversight can result in weak correlation between the extracted features and categories. To address these challenges, we propose a novel model that uses the known information from multiple subjects to bolster EEG classification for an individual subject through adversarial learning strategies. Our method begins by extracting both shallow and attention-driven deep features from EEG signals. Subsequently, we employ a class discriminator to encourage the same-class features from different domains to converge while ensuring that the different-class features diverge. This is achieved using our proposed discrimination loss function, which is designed to minimize the feature distance for samples of the same class across different domains while maximizing it for those from different classes. Additionally, our model incorporates two parallel classifiers that are harmonious yet distinct and jointly contribute to decision-making. Extensive testing on two publicly available EEG datasets has validated our model's efficacy and superiority.

Equitable Skin Disease Prediction Using Transfer Learning and Domain Adaptation

In the realm of dermatology, the complexity of diagnosing skin conditions manually necessitates the expertise of dermatologists. Accurate identification of various skin ailments, ranging from cancer to inflammatory diseases, is paramount. However, existing artificial intelligence (AI) models in dermatology face challenges, particularly in accurately diagnosing diseases across diverse skin tones, with a notable performance gap in darker skin. Additionally, the scarcity of publicly available, unbiased datasets hampers the development of inclusive AI diagnostic tools. To tackle the challenges in accurately predicting skin conditions across diverse skin tones, we employ a transfer-learning approach that capitalizes on the rich, transferable knowledge from various image domains. Our method integrates multiple pre-trained models from a wide range of sources, including general and specific medical images, to improve the robustness and inclusiveness of the skin condition predictions. We rigorously evaluated the effectiveness of these models using the Diverse Dermatology Images (DDI) dataset, which uniquely encompasses both underrepresented and common skin tones, making it an ideal benchmark for assessing our approach. Among all methods, Med-ViT emerged as the top performer due to its comprehensive feature representation learned from diverse image sources. To further enhance performance, we conducted domain adaptation using additional skin image datasets such as HAM10000. This adaptation significantly improved model performance across all models.

Knowledge Graph Modeling-Driven Large Language Model Operating System (LLM OS) for Task Automation in Process Engineering Problem-Solving

We present the Process Engineering Operations Assistant (PEOA), an AI-driven framework designed to solve complex problems in the chemical and process industries. The framework employs a modular architecture orchestrated by a meta-agent, which serves as the central coordinator, managing an action generator and instruction-tuned small-scale language models (expert models). The action generator decomposes complex problems into sub-tasks and identifies suitable expert models to execute each, delivering precise solutions for multi-step problem-solving. Key techniques include advanced knowledge modeling using property graphs for improved information retrieval, facilitating more accurate and contextually relevant solutions. Additionally, the framework utilizes a teacher-student transfer-learning approach with GPT-4 (Omni) to fine-tune the action generator and expert models for domain adaptation, alongside an iterative problem-solving mechanism with sophisticated error handling. Custom datasets were developed to evaluate the framework against leading proprietary language models on various engineering tasks. The results demonstrate the framework’s effectiveness in automating calculations, accelerating prototyping, and providing AI-augmented decision support for industrial processes, marking a significant advancement in process engineering capabilities.

A single-layer dense convolutional reversible residual network for bearing fault diagnosis based on differential local adaptive

Abstract For the cross-domain fault diagnosis of industrial bearings under different working conditions and noise, most current domain adaptation methods in transfer learning only focus on either marginal distribution alignment or conditional distribution alignment. They fail to adequately combine discriminative and global distribution information. Furthermore, the majority of models have a very high parameter count and memory utilization, which makes it challenging to use them in real-world industrial situations. Therefore, a single-layer densely connected reversible residual network based on differential local adaptation is proposed. This network is more competitive in industrial applications than other fault diagnosis models since it not only uses less memory and has fewer parameters, but it also shows superior cross-domain fault diagnostic capacity in noisy situations. Additionally, to extract discriminative and global domain-invariant features, a domain adaptation module is created that takes into account local and global data distributions differently. Multiple transfer tasks and two distinct datasets are used to validate the model. Comparative tests reveal that the suggested model uses less memory and requires fewer parameters to attain good accuracy and transferability.

Day2Dark: Pseudo-Supervised Activity Recognition Beyond Silent Daylight

AbstractThis paper strives to recognize activities in the dark, as well as in the day. We first establish that state-of-the-art activity recognizers are effective during the day, but not trustworthy in the dark. The main causes are the limited availability of labeled dark videos to learn from, as well as the distribution shift towards the lower color contrast at test-time. To compensate for the lack of labeled dark videos, we introduce a pseudo-supervised learning scheme, which utilizes easy to obtain unlabeled and task-irrelevant dark videos to improve an activity recognizer in low light. As the lower color contrast results in visual information loss, we further propose to incorporate the complementary activity information within audio, which is invariant to illumination. Since the usefulness of audio and visual features differs depending on the amount of illumination, we introduce our ‘darkness-adaptive’ audio-visual recognizer. Experiments on EPIC-Kitchens, Kinetics-Sound, and Charades demonstrate our proposals are superior to image enhancement, domain adaptation and alternative audio-visual fusion methods, and can even improve robustness to local darkness caused by occlusions. Project page: https://xiaobai1217.github.io/Day2Dark/.

Intelligent Helicopter Turbine Engine Fault Diagnosis Using Multi-Head Attention

A turbine engine provides power to the helicopter, enabling the helicopter to travel and hover in the air. Since the rotorcraft operates at high altitudes, ensuring safety and maintaining a healthy operational status are crucial at all times. Therefore, a prognostics and health management (PHM) system for the turbine engine must be implemented to predict any anomalies or faults to prevent catastrophic accidents. This research proposes a novel fault diagnosis method for helicopter turbine engines based on operational data acquired from actual aircraft. First, the proposed method predicts engine torque using other operational data while accounting for uncertainty. A Bayesian regression approach is employed to predict the engine torque. The torque margin, defined as the difference between the actual torque and the estimated torque, is then used to diagnose engine faults. Specifically, a multi-head attention mechanism is incorporated to capture interactions between various engine parameters. Additionally, domain adaptation techniques are applied to enhance the model's generalization performance, ensuring robustness across diverse operating conditions. The proposed method is validated using seven different datasets, each acquired from a helicopter engine. Four datasets were used for training, while the remaining three were allocated for testing and validation. The results indicated that the proposed method accurately predicted torque. Furthermore, the fault diagnosis showed promising results, leading to a 3rd-place finish in the 2024 PHM Society Data Challenge in terms of validation score.

Vision transformers for glioma classification using T1 magnetic resonance imaging

Automated image analysis and classification have increasingly advanced in recent decades owing to machine learning and computer vision. In particular, deep learning (DL) architectures have become popular in resource-limited and labor-restricted environments such as the health-care sector. Transformer architecture, a DL method with self-attention mechanism, excels in natural language processing; however, its application in image-based diagnosis in health-care sector remains limited. Herein, the feasibility, bottlenecks, and performance of transformers in magnetic resonance imaging (MRI)-based brain tumor classification were investigated. To this end, a vision transformer (ViT) model was trained and tested using the popular Brain Tumor Segmentation (BraTS) 2015 dataset for glioma classification. Owing to limited data availability, domain adaptation techniques were used to pretrain the ViT model and the BraTS 2015 dataset was used for its fine-tuning. With the model only trained for 100 epochs, the confusion matrix for the two-class problem of tumor and nontumor classification showed an overall classification accuracy of 81.8%. In conclusion, although convolutional neural networks are traditionally used for DL-based medical image classification owing to their attention mechanism and long-range dependency-capturing capability, ViTs can outperform them in MRI-based brain tumor classification.

Contrastive learning-enabled digital twin framework for fault diagnosis of rolling bearing

Abstract Rolling bearings are essential components in various industrial machines, and their failures can lead to significant downtime and maintenance costs. Traditional data-driven fault diagnosis methods often require extensive fault datasets for training, which may not always be available in critical industrial scenarios, limiting their practicality. Digital twins, virtual representations of physical entities reflecting their operational conditions, offer a promising solution for the fault diagnosis of rolling bearings with limited fault data. In this paper, we propose a novel digital twin-driven framework to address the challenge of limited training data in rolling bearing fault diagnosis. Firstly, a virtual bearing simulation model is used to generate the simulated data. Subsequently, a transformer-based network is introduced to learn the discrepancy features from the raw data. Then, a Maximum Mean Discrepancy loss and a supervised contrastive learning loss for raw and augmentation data are established to achieve global domain alignment and instance-based domain alignment. Finally, an unsupervised contrastive learning loss for the augmentation data of the target domain is established to further improve the diagnostic performance. In five cross-domain fault diagnosis tasks representing real industrial scenarios set in this study, the proposed method achieved an average diagnostic accuracy of 84.40%, exceeding two existing advanced domain adaptation methods by more than 10%. The experimental results demonstrate that the proposed method achieves high diagnostic performance in real industrial scenarios where labeled data is lacking. This shows its significant benefits for monitoring the condition of critical bearings.

Forget to Learn (F2L): Circumventing plasticity–stability trade-off in continuous unsupervised domain adaptation

In continuous unsupervised domain adaptation (CUDA), deep learning models struggle with the stability-plasticity trade-off—where the model must forget old knowledge to acquire new one. This paper introduces the “Forget to Learn” (F2L), a novel framework that circumvents such a trade-off. In contrast to state-of-the-art methods that aim to balance the two conflicting objectives, stability and plasticity, F2L utilizes active forgetting and knowledge distillation to circumvent the conflict’s root causes. In F2L, dual-encoders are trained, where the first encoder – the ‘Specialist’ – is designed to actively forget, thereby boosting adaptability (i.e., plasticity) and generating high-accuracy pseudo labels on the new domains. Such pseudo labels are then used to transfer/accumulate the specialist knowledge to the second encoder—the ‘Generalist’ through conflict-free knowledge distillation. Empirical and ablation studies confirmed F2L’s superiority on different datasets and against different SOTAs. Furthermore, F2L minimizes the need for hyperparameter tuning, enhances computational and sample efficiency, and excels in problems with long domain sequences—key advantages for practical systems constrained by hardware limitations.

Knowledge-based inductive bias and domain adaptation for cell type annotation.

Measurement techniques often result in domain gaps among batches of cellular data from a specific modality. The effectiveness of cross-batch annotation methods is influenced by inductive bias, which refers to a set of assumptions that describe the behavior of model predictions. Different annotation methods possess distinct inductive biases, leading to varying degrees of generalizability and interpretability. Given that certain cell types exhibit unique functional patterns, we hypothesize that the inductive biases of cell annotation methods should align with these biological patterns to produce meaningful predictions. In this study, we propose KIDA, Knowledge-based Inductive bias and Domain Adaptation. The knowledge-based inductive bias constrains the prediction rules learned from the reference dataset, composed of multiple batches, to functional patterns relevant to biology, thereby enhancing the generalization of the model to unseen batches. Since the query dataset also contains gaps from multiple batches, KIDA's domain adaptation employs pseudo labels for self-knowledge distillation, effectively narrowing the distribution gap between model predictions and the query dataset. Benchmark experiments demonstrate that KIDA is capable of achieving accurate cross-batch cell type annotation.

CycleGAN-based Data Augmentation for Enhanced Remaining Useful Life Prediction Under Unsupervised Domain Adaptation

Predictive maintenance is crucial for enhancing operational efficiency and reducing costs in Prognostics and Health Management (PHM). One of the key tasks in predictive maintenance is the estimation of Remaining Useful Life (RUL) of machinery. In practice, the data for different machines is not always accessible in sufficient quantity or quality, therefore the machine learning models trained on machines in one domain often perform poorly when applied to other domains due to covariate shifts. As a solution, Domain Adaptation (DA) aims to tackle domain shifts by extracting domain-invariant features. However, traditional methods often fail to adequately address the complexity and variability of real-world data. We propose to address this challenge, using a Wasserstein CycleGAN with Gradient Penalty (W-CycleGAN-GP) to learn mappings between domains and generate augmented data in the target domain from data in the source domain. We use our approach to generate realistic augmented data that bridge domain gap coupled with recent work on adversarial-based and correlation alignment-based DA models to improve the performance of RUL prediction models in target domains without having access to labeled data. The experimental results on the C-MAPSS dataset demonstrate a significant improvement in the RUL prediction score and accuracy within the target domain.

Damage Diagnostics of Miter Gates Using Domain Adaptation and Normalizing Flow-Based Likelihood-Free Inference

Miter gates are vital civil infrastructure components in inland waterway transportation networks. To provide risk-informed insights for decisions related to repair and maintenance, sensors have been installed on some miter gates for monitoring. Despite the monitoring system's ability in collecting a large volume of monitoring data, accurately diagnosing damage state in such large structures remains challenging due to the lack of labeled monitoring data, since these structures are designed with high reliability and for a long operation life. This paper addresses this challenge by proposing a damage diagnostics approach for miter gates based on domain adaptation. The proposed approach consists of two main modules. In the first module, Cycle-Consistent generative adversarial network (CycleGAN) is employed to map monitoring data of a miter gate of interest and other similar yet different miter gates into the same analysis domain. Subsequently, a normalizing flow-based likelihood-free inference model is constructed within this common domain using data from source miter gates whose damage states are labeled from historical inspections. The trained normalizing flow model is then used to predict the damage state of the target miter gate based on the translated monitoring data. A case study is presented to demonstrate the effectiveness of the proposed method. The results indicate that the proposed method in general can accurately estimate the damage state of the target miter gate in the presence of uncertainty.

Multidisciplinary Framework for Vibration-Based Gear Fault Diagnosis Using Experiments, Modeling, and Machine Learning

Vibration-based gear diagnosis is crucial for ensuring the reliability of rotating machinery, making the monitoring of gear health essential for preventing costly downtime and optimizing performance. This study proposes a multidisciplinary framework to enhance gear diagnosis, that aligns with the new era of digital twins by integrating experiments, dynamic modeling, physical preprocessing, and machine learning. Within this framework, we focus on three core procedures: domain adaptation to reduce discrepancies between measured data and synthetic data generated by dynamic models; physical preprocessing, grounded in in-depth investigations dictating signal processing and feature engineering techniques; and learning algorithms, encompassing the process of training AI-based models. We demonstrate this framework through a comprehensive case study of localized tooth fault diagnosis, using controlled-degradation tests and realistic simulations. First, we detect faults using unsupervised learning algorithms; then, we use zero-shot-learning for classifying between localized and distributed faults; finally, we adopt a one-shot-learning strategy for severity estimation. Above all, this hybrid framework bridges the gaps between physical-based and AI-based approaches by combining physical knowledge and advanced algorithmics with machine learning. This contributes to the PHM field by offering valuable insights into integrating different aspects of research, thereby enhancing performance in gear diagnosis tasks.