Domain Adaptation Research Articles

A GraphKAN-Based Intelligent Fault Diagnosis Method of Rolling Bearing Under Variable Working Conditions

Unsupervised domain adaptation (UDA) can effectively address the two main drawbacks of transfer learning: the requirement of a large number of samples collected from different working conditions, and the inherent defects of convolutional neural networks (CNNs). In the realm of UDA, it is essential to leverage three types of information: class labels, domain specifications, and data organization. These components play a vital role in linking the source domain with the target domain. A technique aimed at identifying issues in rolling bearings is presented, employing an integration of CNN-KAN and GraphKAN structures to support the UDA methodology. A cohesive deep learning architecture is employed to represent the three types of information involved in UDA. The initial two types of information are represented through the roles of classifier and domain discriminator. To begin with, an architecture leveraging CNN-KAN is employed to extract features from the incoming signals. Following this, the features obtained from the CNN-KAN architecture are input into a specially developed graph creation layer that constructs instance graphs by analyzing the relationships among the structural characteristics found within the samples. In the following step, an innovative GraphKAN model is applied to illustrate the instance graphs, concurrently employing CORrelation ALignment (CORAL) loss to assess the structural discrepancies among instance graphs from different domains. Results from experiments conducted on two separate datasets demonstrate that the proposed framework surpasses alternative approaches and successfully recognizes transferable characteristics that are advantageous for domain adaptation.

Unified Text-Image Space Alignment with Cross-Modal Prompting in CLIP for UDA

Unsupervised Domain Adaptation (UDA) aims to transfer models trained on a labeled source domain to an unlabeled target domain. Due to the excellent generalization ability of Vision Language Models (VLMs) such as CLIP in downstream tasks, most recent methods apply CLIP to UDA tasks through learning domain-specific text prompts for source and target domains separately. However, these methods fail to dynamically adjust image features based on the characteristics of their respective domains, thereby limiting their alignment with domain-specific text prompts in CLIP’s joint space, which is a key factor in improving classification performance in the target domain. To bridge this gap, we propose a Unified Text-Image Space Alignment with Cross-Modal Prompting (UTISA) framework for UDA. Firstly, we introduce a cross-modal prompt learning (CMP) module to generate domain-specific image prompts and layer-specific image prompts for the visual branch to encode domain-specific knowledge globally and locally. Secondly, under the guidance of image prompts, we introduce a domain-aware multi-layer feature fusion (DMF) module to construct multi-layer domain features for each domain and enhance the image features with these multi-layer domain features, which enables the image features to better reflect the characteristics of their respective domains, thereby promoting their alignment with domain-specific text prompts. Moreover, we introduce a perturbation-driven regularization (PDR) mechanism for the target domain to enhance the robustness and generalization of the model. The experiments demonstrate that UTISA achieves the best performance on three mainstream UDA benchmarks, including 87.9% on OfficeHome, 90.9% on VisDA-2017, 62.4% on DomainNet.

Online and Cross-User Finger Movement Pattern Recognition by Decoding Neural Drive Information from Surface Electromyogram.

Cross-user variability is a well-known challenge that leads to severe performance degradation and impacts the robustness of practical myoelectric control systems. To address this issue, a novel method for myoelectric recognition of finger movement patterns is proposed by incorporating a neural decoding approach with unsupervised domain adaption (UDA) learning. In our method, the neural decoding approach is implemented by extracting microscopic features characterizing individual motor unit (MU) activities obtained from a two-stage online surface electromyogram (SEMG) decomposition. A specific deep learning model is designed and initially trained using labeled data from a set of existing users. The model can update adaptively when recognizing the movement patterns of a new user. The final movement pattern was determined by a fuzzy weighted decision strategy. SEMG signals were collected from the finger extensor muscles of 15 subjects to detect seven dexterous finger-movement patterns. The proposed method achieved a movement pattern recognition accuracy of ([Formula: see text])% over seven movements under cross-user testing scenarios, much higher than that of the conventional methods using global SEMG features. Our study presents a novel robust myoelectric pattern recognition approach at a fine-grained MU level, with wide applications in neural interface and prosthesis control.

Data Augmentation Techniques Using Generative Adversarial Networks: Employing GANs to Create Synthetic Data for Enhancing Machine Learning Model Training

The capability of GANs to increase data size has been a revelation to people in the augmentation field. It has saved machine learning algorithms with scarcity, imbalance, or poor data variation. Unlike the previous approaches based on transformation techniques, GANs can learn complex patterns and generate realistic synthetic data, which have virtually never been possible for fields and applications. In this paper, 12 new GAN-augmentation strategies are proposed and implemented. These are image-to-image translation, text-to-image synthesis, style transfer, and domain adaptation. These techniques prove that using GANs can enhance the quality of the data and models, balance the data, and maintain the confidentiality of the data while applying any of these techniques. The use of GANs in generating new data to enrich data sets has proved promising, especially in specific fields like health, where medical images complement diagnostic models while respecting patients' rights to privacy. It has also been used in image recognition when it creates diverse outputs to solve problems such as class imbalance, besides solving regular problems faced in natural language processing, including text-to-image synthesis. Compared to previous approaches, GANs provide better and more complex solutions and fill the data with newcomers instead of transforming the given ones. This capability leads to the creation of much more diverse datasets that can be closer to the real environment. Nevertheless, there are some key issues of concern, such as computational cost, mode collapse, and, most notably, social impact issues, where GANs may be misused using deepfake technologies. The proposed techniques show great potential and are based on a new approach to using data augmentation to solve modern tasks. However, the lack of detailed experiments and decreased accuracy compared to machine learning algorithms emphasize the need for future research to confirm the efficiency of the proposed approach in various applications. Overcoming these limitations via solid architectures, appropriate measures of performance, and policies is crucial. While GANs are still a relatively young type of AI technology, their enhancement and scaling can create significant opportunities for using these models in combination with other complex models, such as diffusion and reinforcement learning.

IF-EDAAN: An information fusion-enhanced domain adaptation attention network for unsupervised transfer fault diagnosis

Unlike domain adaptation methods that rely on single-source information for transfer diagnosis, multi-source information-based domain adaptation methods can leverage the extensive diagnostic features derived from multiple sources of data. However, the issues of potential feature conflicts, the critical fault information loss, and high computational burdens still hinder effective applications of multi-source information domain adaptation for transfer diagnosis. For resolving these issues, this study proposes an unsupervised multi-source information domain adaptation approach for transfer fault diagnosis, which utilizes an information fusion-enhanced domain adaptation attention network (IF-EDAAN). Firstly, an information fusion method that converts multi-source information into a fused image using principal component analysis and signal-to-image conversion is employed to enhance and compress data from both the source and target domains. Then, a parameter-free attention mechanism (PFAM) module is proposed to adaptively focus on the domain-invariant temporal and spatial features of information fusion samples. Subsequently, the weight assignment module and joint maximum mean discrepancy metric strategy are proposed to mitigate negative transfer, thus enabling the effective extraction and alignment of domain-invariant temporal and spatial features. Finally, experiment validations on two rotating machinery datasets have been comprehensively elaborated to verify the efficacy and advantages of our proposed IF-EDAAN approach for transfer fault diagnosis across different working conditions. Experiment results have proved that our proposed IF-EDAAN approach can rapidly adapt to new transfer diagnostic scenarios with impressive performance and outperform several mainstream unsupervised domain adaptation approaches.

Investigating a Domain Adaptation Approach for Integrating Different Measurement Instruments in a Longitudinal Clinical Registry.

In a longitudinal clinical registry, different measurement instruments might have been used for assessing individuals at different time points. To combine them, we investigate deep learning techniques for obtaining a joint latent representation, to which the items of different measurement instruments are mapped. This corresponds to domain adaptation, an established concept in computer science for image data. Using the proposed approach as an example, we evaluate the potential of domain adaptation in a longitudinal cohort setting with a rather small number of time points, motivated by an application with different motor function measurement instruments in a registry of spinal muscular atrophy (SMA) patients. There, we model trajectories in the latent representation by ordinary differential equations(ODEs), where person-specific ODE parameters are inferred from baseline characteristics. The goodness of fit and complexity of the ODE solutions then allow to judge the measurement instrument mappings. We subsequently explore how alignment can be improved by incorporating corresponding penalty terms into model fitting. To systematically investigate the effect of differences between measurement instruments, we consider several scenarios based on modified SMA data, including scenarios where a mapping should be feasible in principle and scenarios where no perfect mapping is available. While misalignment increases in more complex scenarios, some structure is still recovered, even if the availability of measurement instruments depends on patient state. A reasonable mapping is feasible also in the more complex real SMA data set. These results indicate that domain adaptation might be more generally useful in statistical modeling for longitudinal registrydata.

A partial domain adaptation broad learning system for machinery fault diagnosis

A partial domain adaptation broad learning system for machinery fault diagnosis

Cross-domain damage detection through partial conditional adversarial domain adaptation

Cross-domain damage detection through partial conditional adversarial domain adaptation

A novel two-dimensional progressive domain adaptation framework for cross-domain remaining useful life prediction

A novel two-dimensional progressive domain adaptation framework for cross-domain remaining useful life prediction

Domain knowledge boosted adaptation: Leveraging vision-language models for multi-source domain adaptation

Domain knowledge boosted adaptation: Leveraging vision-language models for multi-source domain adaptation

Pseudo-class distribution guided multi-view unsupervised domain adaptation for hyperspectral image classification

Pseudo-class distribution guided multi-view unsupervised domain adaptation for hyperspectral image classification

Enabling efficient cross-building HVAC fault inferences through novel unsupervised domain adaptation methods

Enabling efficient cross-building HVAC fault inferences through novel unsupervised domain adaptation methods

Towards Enhanced Interpretability: A Mechanism-Driven domain adaptation model for bearing fault diagnosis across operating conditions

Towards Enhanced Interpretability: A Mechanism-Driven domain adaptation model for bearing fault diagnosis across operating conditions

Semantic-aware for point cloud domain adaptation with self-distillation learning

Semantic-aware for point cloud domain adaptation with self-distillation learning

A dynamic weighted joint distribution domain adaptation network for cross-machine fault diagnosis of rolling bearings

A dynamic weighted joint distribution domain adaptation network for cross-machine fault diagnosis of rolling bearings

Causality-driven feature selection and domain adaptation for enhancing chemical foundation models in downstream tasks

Abstract Recent advancements in large foundation models have revealed impressive capabilities in mastering complex chemical language representations. These models undergo a task-agnostic learning phase, characterized by pre-training on extensive unlabeled corpora followed by fine-tuning on specific downstream tasks. This methodology reduces reliance on labeled data, facilitating data acquisition and broadening the scope of chemical language representation. However, real-world scenarios often pose challenges due to domain shift, a phenomenon where the data distribution in downstream tasks differs from that of the pre-training phase, potentially degrading model performance. To address this, we present a novel causal-based framework for feature selection and domain adaptation to enhance the performance of chemical foundation models on downstream tasks. Our approach employs a multi-stage feature selection method that identifies physico-chemical features based on their direct causal-effect over specific downstream properties. By employing Mordred descriptors and Markov blanket causal graphs, our approach provides insight into the causal relationships between features and target properties for prediction tasks. We evaluate our approach on various foundation model architectures and datasets, demonstrating performance improvements, which showcases the robustness and the agnostic nature of our approach.

A deep transfer network based on dual-task learning for predicting the remaining useful life of rolling bearings

Abstract Rolling bearings are essential in rotating machinery, and accurate remaining useful life (RUL) predictions are necessary for effective maintenance and optimal performance. Poor trend of health indicators (HI) and operating condition variations can reduce the reliability and accuracy of RUL predictions.To address these challenges, we propose a deep transfer network based on dual-task learning for predicting the remaining life of rolling bearings (DTLDL-RUL). This framework integrates health status assessment and RUL prediction, leveraging task commonalities and differences to create a robust model, then transfers it to improve generalization in the target domain. Specifically, we first mine spatiotemporal features from vibration signals to generate and classify HI, labeling them for subsequent tasks. Next, we use the shared feature extractors and private residual networks to capture common and specific features of each task, merge them with the multi-gate control networks, and adaptively adjust task weights in the loss function to enhance model adaptability. Finally, the trained model is transferred to the target domain using domain adaptation to extract domain invariant features and consider target-specific features, enhancing generalization. Experiments conducted on the 2012 PHM and XJTU-SY datasets demonstrate that the proposed method achieves high accuracy and generalization in RUL predictions.

Few-Shot Object Detection Based on Global Domain Adaptation Strategy

Aiming to detect novel objects from only a few annotated samples, few-shot object detection (FSOD) has undergone remarkable development. Previous works rarely pay attention to the perspective of gradient propagation to optimize existing methods, therefore failing to make full use of information for novel objects in gradient propagation. We propose a method to solve this problem based on two-stage fine-tuning. A domain adaptation module with multi-constraints is used to promote the spread of gradients, a classification promotion network is used to improve the effect of classification, and a multi-path mask head is added to enrich RoI features. Experiments on PASCAL VOC and COCO datasets show that our model significantly raises the performance compared with previous methods (up to 1–5% in average).

A temporal-spatial feature fusion network for emotion recognition with individual differences reduction.

In the context of EEG-based emotion recognition tasks, a conventional strategy involves the extraction of spatial and temporal features, subsequently fused for emotion prediction. However, due to the pronounced individual variability in EEG and the constrained performance of conventional time-series models, cross-subject experiments often yield suboptimal results. To address this limitation, we propose a novel network named Time-Space Emotion Network (TSEN), which capitalizes on the fusion of spatiotemporal information for emotion recognition. Diverging from prior models that integrate temporal and spatial features, our network introduces a Convolutional Block Attention Module (CBAM) during spatial feature extraction to judiciously allocate weights to feature channels and spatial positions. Furthermore, we bolster network stability and improve domain adaptation through the incorporation of a residual block featuring Switchable Whitening (SW). Temporal feature extraction is accomplished using a Temporal Convolutional Network (TCN), ensuring elevated prediction accuracy while maintaining a lightweight network structure. We conduct experiments on the preprocessed DEAP dataset. Ultimately, the average accuracy for arousal prediction is 0.7032 with a variance of 0.0876, and the F1 score is 0.6843. For valence prediction, the accuracy is 0.6792 with a variance of 0.0853, and the F1 score is 0.6826. TSEN exhibits high accuracy and low variance in cross-subject emotion prediction tasks, effectively reducing individual differences among different subjects. Additionally, TSEN has a smaller parameter count, enabling faster execution.

A new cross-domain approach for bearing fault diagnosis based on multiscale convolutional networks and adversarial subdomain adaptation

ABSTRACT Fault diagnosis of rolling bearings is of paramount significance in the field of engineering, as it directly impacts the reliability and safety of mechanical systems. Although deep learning techniques have demonstrated promising performance in this field, their efficacy often diminishes under varying operational conditions or when labelled data is limited. To overcome these challenges, this paper introduces a more precise cross-domain approach for bearing fault diagnosis. The proposed method exploits the temporal multiscale characteristics of vibration signals and the inherent multiscale nature of bearing faults by constructing a one-dimensional multiscale convolutional neural network. This network extracts features at multiple scales from raw vibration signals, which are then fused to form robust and generalised representations. Additionally, the integration of the Efficient Channel Attention mechanism further refines feature selection, enhancing the overall performance of the model. The label classifier, in conjunction with the Nuclear-norm Wasserstein Discrepancy, serves as a domain discriminator to facilitate adversarial domain adaptation. Concurrently, local maximum mean discrepancy and adversarial domain adaptation techniques align both global and subdomain feature distributions. Furthermore, label smoothing is incorporated to enhance the model’s generalisation capabilities. Experimental validation on the CWRU and PU rotating machinery datasets demonstrates the method’s exceptional robustness and superior transferability.