Domain Generalization Methods Research Articles

Contrastive domain-invariant generalization for remaining useful life prediction under diverse conditions and fault modes

As industrial equipment becomes increasingly complex, necessitating operation under varied conditions and often exhibiting diverse failure modes, traditional deep learning models built on data from the original environment become inapplicable. Moreover, in actual industrial scenarios, the generalization capability of Domain Adaptation and classic Domain Generalization methods is severely impacted when there is a lack of multiple source domain and target domain data, due to the cost or feasibility constraints associated with collecting extensive monitoring data. In this paper, a single domain Contrastive Domain-Invariant Generalization (CDIG) method for estimating the remaining useful life under different conditions and fault modes is proposed. This method first defines homologous signals as the foundational data. Subsequently, it learns domain-invariant features by encouraging two feature extraction processes to extract latent features of homologous signals as similarly as possible. Additionally, multiple condition-based attention, pooling, and a novel equalization loss function are utilized to regulate the generation of domain-invariant features. Ultimately, the RUL predictor is trained by source domain data, operational conditions, and temporal information to facilitate its applicability across diverse domains. Case studies demonstrate that CDIG achieves satisfactory predictive results under unseen conditions, highlighting the potential of the proposed method as an effective predictive tool.

A Patch Diversity Transformer for Domain Generalized Semantic Segmentation.

Domain generalization (DG) is one of the critical issues for deep learning in unknown domains. How to effectively represent domain-invariant context (DIC) is a difficult problem that DG needs to solve. Transformers have shown the potential to learn generalized features, since the powerful ability to learn global context. In this article, a novel method named patch diversity Transformer (PDTrans) is proposed to improve the DG for scene segmentation by learning global multidomain semantic relations. Specifically, patch photometric perturbation (PPP) is proposed to improve the representation of multidomain in the global context information, which helps the Transformer learn the relationship between multiple domains. Besides, patch statistics perturbation (PSP) is proposed to model the feature statistics of patches under different domain shifts, which enables the model to encode domain-invariant semantic features and improve generalization. PPP and PSP can help to diversify the source domain at the patch level and feature level. PDTrans learns context across diverse patches and takes advantage of self-attention to improve DG. Extensive experiments demonstrate the tremendous performance advantages of the PDTrans over state-of-the-art DG methods.

Domain generalization for machine compound fault diagnosis by Domain-Relevant Joint Distribution Alignment

In machine fault diagnosis, domain generalization methods have gained significant attention due to their advantage in non-requirement of priori target domain distribution. However, they pose great challenges in domain-relevant feature learning and incurs inferior unseen-domain classification when large domain divergence exists. To address this issuse, we propose a novel distribution alignment strategy named DRJDA (Domain-Relevant Joint Distribution Alignment) that matches the domain-joint distribution and domain-relevant distribution for domain generalization fault diagnosis. Specifically, the α-PE divergence is employed to minimize the distribution discrepancy, which is demonstrated to be explicitly derived as the maximum value of a quadratic function. Additionally, a parameter-free plug-and-play data augmentation module that performs feature-level instance mixture and style transfer to increase the generalization ability. Finally, data from the China Light-duty Vehicle Test Cycle (CLTC) tests are used as case studies, and the experiments carried out across 14 different domains prove the proficiency of the proposed DRJDA in learning domain-invariant feature when significant domain divergence exists, indicating its remarkable potential in compound fault diagnosis for industrial machinery.

A Partial Domain Generalization Method for Modeling Multiple Multistage Manufacturing Processes

In complex manufacturing systems, materials are processed by machines sequentially before final products are get, forming a multistage manufacturing process (MMP). In a modern massive production factory, it is common that multiple MMPs work simultaneously to fulfill productivity needs. Multiple MMPs may contain machines with different running time, processing the same types of products but with different specifications. In such a Multiple MMP(MMMP) system, it is critical for machines to learn from each other to gain optimal parameter settings of each. However, traditional machine learning methods usually fail to consider both similarities and unique features among multiple processes at the same time. To address this problem, a partial domain generalization method is proposed based on the thought of transfer learning to combine useful information from historical processes while maintaining features from the in-production process. The proposed method also suits for problems with the assumption that only partial input variables are available. Studies reveal that the proposed method has superior prediction performance over traditional machine learning methods and some widely-used transfer learning methods.

Bilateral Proxy Federated Domain Generalization for Privacy-preserving Medical Image Diagnosis.

Contemporary domain generalization methods have demonstrated effectiveness in aiding the generalized diagnosis of medical images with multi-source data by joint optimization. However, the centralized training paradigm employed by these approaches becomes infeasible when data are non-shared across domains due to the high privacy of medical data. Despite attempts by existing federated domain generalization methods to address this issue, the simultaneous attainment of strict privacy protection and a satisfactory level of generalization ability on out-of-distribution data remains a persistent challenge. In this paper, to tackle this challenging problem, we propose a novel approach called the Bilateral Proxy Framework (BPF). The BPF leverages the client-side proxies to facilitate the strict privacy-preserving communications with the server and ensure smoother and more stable convergences of local models through mutual distillation. Meanwhile, the server-side proxy adopts a distance-based strategy and a parameter moving average scheme, which enhances the stability and robustness of the global model, particularly by averting abrupt parameter changes that could result in fluctuations or overfitting. Through these advancements, our framework strives to enhance the generalization capability of the global model, enabling more accurate and reliable medical image diagnosis in federated settings. The effectiveness of our method is demonstrated with superior performance over state-of-the-arts on both simulated and real-world distribution medical image diagnosis tasks.

SiMix: A domain generalization method for cross-site brain MRI harmonization via site mixing

Brain magnetic resonance imaging (MRI) is widely used in clinical practice for disease diagnosis. However, MRI scans acquired at different sites can have different appearances due to the difference in the hardware, pulse sequence, and imaging parameter. It is important to reduce or eliminate such cross-site variations with brain MRI harmonization so that downstream image processing and analysis is performed consistently. Previous works on the harmonization problem require the data acquired from the sites of interest for model training. But in real-world scenarios there can be test data from a new site of interest after the model is trained, and training data from the new site is unavailable when the model is trained. In this case, previous methods cannot optimally handle the test data from the new unseen site. To address the problem, in this work we explore domain generalization for brain MRI harmonization and propose Site Mix (SiMix). We assume that images of travelling subjects are acquired at a few existing sites for model training. To allow the training data to better represent the test data from unseen sites, we first propose to mix the training images belonging to different sites stochastically, which substantially increases the diversity of the training data while preserving the authenticity of the mixed training images. Second, at test time, when a test image from an unseen site is given, we propose a multiview strategy that perturbs the test image with preserved authenticity and ensembles the harmonization results of the perturbed images for improved harmonization quality. To validate SiMix, we performed experiments on the publicly available SRPBS dataset and MUSHAC dataset that comprised brain MRI acquired at nine and two different sites, respectively. The results indicate that SiMix improves brain MRI harmonization for unseen sites, and it is also beneficial to the harmonization of existing sites.

Simulation data-driven attention fusion network with multi-similarity metric: A single-domain generalization diagnostic method for tie rod bolt loosening of a rod-fastening rotor system

The rod-fastening rotor system is the core rotating component of gas turbines and aero engines. The rod-disk joint structure of the rod-fastening rotor system is subjected to tie rod bolt loosening under complex operating conditions. Because of the uncertainty of fault occurrence, the model must possess diagnostic capability under unseen working conditions. Domain generalization methods are commonly employed to address such problems. They typically utilize multi-condition data or generate data from a single condition to train the model, enabling it to generalize to unknown operating conditions. However, difficulties in collecting complete multi-condition data and limitations in generating data, such as conflicting attributes and limited categories, make it challenging to ensure high-quality data for model training. To address these issues, this paper establishes a dynamic model of the rod-fastening rotor system to obtain high-fidelity simulation fault data, ensuring the physical accuracy, richness, and diversity of the source domain data. Meanwhile, a novel signal-domain generalization method called simulation data-driven attention fusion network with multi-similarity metric is proposed for tie rod bolt loosening in the rod-fastening rotor system. The attention feature fusion strategy is utilized to achieve the comprehensive representation of local and global features efficiently. Moreover, a multi-similarity loss based on deep metric learning is introduced to comprehensively consider the self-similarity and relative similarities between features, enabling class-level optimization of features. The proposed method is validated to have superior diagnostic performance in two cross-domain tasks compared with other methods. The proposed method can provide valuable insights into fault diagnosis in the rod-fastening rotor system.

Domain generalization via causal fine-grained feature decomposition and learning

Domain generalization aims to accurately predict unknown data using models trained by known domain data. Learning domain-invariant representations based on causal inference is one of the popular directions in domain generalization. However, existing domain generalization models based on causal inference cannot correctly determine the invariance conditions of the data. It is because of the interdependent relationship between domain-invariant features and confounding factors that ultimately leads to difficulty in obtaining truly invariant representations by the model. In this paper, we propose a novel domain generalization method called causal fine-grained feature decomposition and learning (CFFDL), which aims to eliminate latent confounding factors and learn causal domain-invariant representations in image classification tasks. Specifically, we design a feature decomposition module based on mutual information, which can decompose deep features into fine-grained feature factors and achieve factor independence between different dimensions, thereby helping to eliminate confounding factors. Furthermore, we introduce a causal representations learning module that can effectively filter and extract relevant causal features of the prediction task while eliminating the influence of confounding factors, improving the model’s performance on domain generalization. Extensive experiments on three domain generalization datasets VLCS, PACS and Office–Home show that our method outperforms the current state-of-the-art models, proving its effectiveness and superiority on domain generalization tasks of image classification.

Towards reliable domain generalization: Insights from the PF2HC benchmark and dynamic evaluations

Deep neural networks (DNNs) are easily biased towards the training set, which causes substantial performance degradation for out-of-distribution data. Many methods are studied to generalize under various distribution shifts in the literature of domain generalization (DG). To facilitate practical DG research, we construct a large-scale non-independent and identically distributed Chinese characters dataset called PaHCC (Printed and Handwritten Chinese Characters) for a real application scenario (generalization from Printed Fonts to Handwritten Characters, PF2HC) of DG methods. We evaluate eighteen DG methods on the proposed PaHCC dataset and demonstrate that the performance of the current algorithms on this dataset remains inadequate. To improve the performance, we propose a radical-based multi-label learning method by integrating structure learning into statistical methods. Moreover, in the dynamic evaluation settings, we discover additional properties of DG methods and demonstrate that many algorithms suffer from unstable performances. We advocate that researchers in the DG community pay attention not only to accuracy under the fixed leave-one-domain-out protocol but also to algorithmic stability across variable training domains in future studies. Our dataset, method, and evaluations bring valuable insights to the DG community and could promote the development of realistic and stable algorithms.

Inter-participant transfer learning with attention based domain adversarial training for P300 detection

A Brain-computer interface (BCI) system establishes a novel communication channel between the human brain and a computer. Most event related potential-based BCI applications make use of decoding models, which requires training. This training process is often time-consuming and inconvenient for new users. In recent years, deep learning models, especially participant-independent models, have garnered significant attention in the domain of ERP classification. However, individual differences in EEG signals hamper model generalization, as the ERP component and other aspects of the EEG signal vary across participants, even when they are exposed to the same stimuli. This paper proposes a novel One-source domain transfer learning method based Attention Domain Adversarial Neural Network (OADANN) to mitigate data distribution discrepancies for cross-participant classification tasks. We train and validate our proposed model on both a publicly available OpenBMI dataset and a Self-collected dataset, employing a leave one participant out cross validation scheme. Experimental results demonstrate that the proposed OADANN method achieves the highest and most robust classification performance and exhibits significant improvements when compared to baseline methods (CNN, EEGNet, ShallowNet, DeepCovNet) and domain generalization methods (ERM, Mixup, and Groupdro). These findings underscore the efficacy of our proposed method.

An adversarial-based domain generalization method for the health evaluation of axial piston pumps

Abstract The axial piston pump is the power component in hydraulic systems and evaluating its health status is of great importance to the safe operation of hydraulic systems. Discharge pressure signals are common monitoring signals for axial piston pumps, but it is difficult to obtain satisfactory health evaluation results by directly using raw discharge pressure signals since the degradation information lies in some specific frequency bands. Furthermore, the axial piston pump often operates under different operating conditions and most existing deep learning models have low diagnostic accuracies due to the problems of insufficient degradation data and different data distribution. This paper proposes an adversarial-based domain generalization (DG) method by integrating time-frequency analysis and data augmentation, to accurately predict the health status of axial piston pumps under unknown working conditions. First, discharge pressure signals under various operating conditions are decomposed by using the complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), and effective intrinsic mode functions (IMFs) are selected to train multiple convolutional neural network (CNN) models. Second, a novel data augmentation method based on the modified discrete cosine transform-composite spectrum (DCS) algorithm is introduced to fuse the IMFs from different domains and generate pseudo data sets. Finally, the adversarial training is adopted between the real data and the pseudo data to capture domain-generalized features. The discharge pressure signal of an actual axial piston pump at different health levels were collected on a test bench to verify the effectiveness of the proposed method. Results indicate that the proposed method has a higher prediction accuracy than the comparative methods.

Causality-inspired multi-source domain generalization method for intelligent fault diagnosis under unknown operating conditions

Currently, fault diagnosis methods based on domain generalization have received widespread attention due to their advantages of not requiring target domain data. Therefore, in this paper, A multi-source domain generalization fault diagnosis method is proposed, consisting of feature diversity activation and non-causal feature suppression from a causal perspective. In the first part, a 3D-Dynamic convolution-based residual network is designed to adaptively learn task related features from different source domains, encouraging the model to focus more on causal feature learning. Furthermore, based on the maximum entropy idea, channel attention diversification is proposed to activate more potential causal features. In the second part, a feature suppression method based on domain discriminator guidance is proposed to explicitly discard non-causal features, specifically, the domain discriminator progressively locates and distinguishes between causal and non-causal features at the layer and channel level and creates binary mask matrices to suppress non-causal related features. Experiments are conducted on the PU and SDUST bearing datasets, and the proposed method can productively solve the cross-domain diagnosis problem under unknown operating conditions.

Domain generalization for semantic segmentation: a survey

Deep neural networks (DNNs) have proven explicit contributions in making autonomous driving cars and related tasks such as semantic segmentation, motion tracking, object detection, sensor fusion, and planning. However, in challenging situations, DNNs are not generalizable because of the inherent domain shift due to the nature of training under the i.i.d. assumption. The goal of semantic segmentation is to preserve information from a given image into multiple meaningful categories for visual understanding. Particularly for semantic segmentation, pixel-wise annotation is extremely costly and not always feasible. Domain generalization for semantic segmentation aims to learn pixel-level semantic labels from multiple source domains and generalize to predict pixel-level semantic labels on multiple unseen target domains. In this survey, for the first time, we present a comprehensive review of DG for semantic segmentation. we present a comprehensive summary of recent works related to domain generalization in semantic segmentation, which establishes the importance of generalizing to new environments of segmentation models. Although domain adaptation has gained more attention in segmentation tasks than domain generalization, it is still worth unveiling new trends that are adopted from domain generalization methods in semantic segmentation. We cover most of the recent and dominant DG methods in the context of semantic segmentation and also provide some other related applications. We conclude this survey by highlighting the future directions in this area.

Domain generalized open-set intelligent fault diagnosis based on feature disentanglement meta-learning

Abstract Existing domain generalization-based intelligent fault diagnosis methods mainly focus on learning domain-invariant features. However, in practical scenarios, these features are difficult to extract and effectively distinguish from class-related features. Moreover, these methods often assume identical label distributions between the source and target domain, making it challenging to handle scenarios where unknown classes exist in the target domain. To address these issues, this paper proposes a domain generalized open-set intelligent fault diagnosis method based on feature disentanglement meta-learning. A binary mask feature disentanglement module is constructed to overcome the information loss caused by feature reconstruction, enabling the separation of domain-specific and class-related features. Additionally, a meta-purification loss function is defined, incorporating a correlation loss term to remove impurity features from the class-related features, and further purifying class information through feature combination pairing. The method is trained on multiple source domains using a meta-learning strategy and generalized to target domains with unknown classes. The method is utilized for bearing fault diagnosis, designing multi-task experimental scenarios under different rotational speeds, and compared with existing domain generalization methods. Experimental results show that the proposed method exhibits excellent generalization ability and effectively addresses the issue of domain generalized open-set fault diagnosis.

Domain composition and attention network trained with synthesized unlabeled images for generalizable medical image segmentation

Despite that deep learning models have achieved remarkable performance in medical image segmentation, their performance is often limited on testing images from new centers with a domain shift. To achieve Domain Generalization (DG) for medical image segmentation, we propose a Domain Composition and Attention-based Network (DCA-Net) combined with structure- and style-based data augmentation that generates unlabeled synthetic images for training. First, DCA-Net represents features in one certain domain by a linear combination of a set of basis representations that are learned by parallel domain preceptors with a divergence constraint. The linear combination is used to calibrate the feature maps of an input image, which enables the model to generalize to unseen domains. Second, considering the number of domains and images for training is limited, we employ generative models to synthesize images with a higher structure diversity, and to leverage the unlabeled synthetic images, we introduce a consistency constraint for their predictions under style augmentation based on frequency amplitude mixture. Additionally, a Test-Time Frequency Augmentation (TTFA) is proposed to neutralize the domain shift from the target to source domains. Experimental results on two multi-domain datasets for fundus structure and nasopharyngeal carcinoma segmentation showed that: (1) our method significantly outperformed several existing DG methods, and (2) the model’s generalizability was largely improved by domain composition and attention modules; (3) by leveraging the unlabeled synthetic images and the TTFA, the model could better deal with images from unseen domains.

Restoration towards decomposition: A simple approach for domain generalization

Domain generalization (DG) aims to train a model capable of generalizing to unseen target domains by utilizing data from multiple disjoint source domains. Previous domain generalization methods primarily utilize data augmentation and domain-invariant feature learning. However, since the target domain is unknown, data augmentation methods struggle to generate images or features that closely resemble the target domain. In addition, existing domain-invariant feature learning methods are incapable of achieving complete decoupling, resulting in misalignment of feature distribution between the source and target domains. Hence, in this work, we propose a new perspective to address domain generalization by focusing on learning the ability to transform the target domain visual representations into those of the source domain. If the style of the unknown target domain can be transformed into the style of the known source domain, the adverse effects of domain shift can be effectively mitigated. Following this line of thought, we have developed a meta-learning task named “feature restoration” aimed at training a model to acquire this capability. Specifically, feature restoration involves replacing the domain-specific components within the visual representations of the target domain with those of the source domain. Experimental results on five DG benchmarks reveal that our method can achieve state-of-the-art performance. Ablation studies and visualization results further demonstrate the rationality and effectiveness of our design.

Unbiased Semantic Representation Learning Based on Causal Disentanglement for Domain Generalization

Domain generalization primarily mitigates domain shift among multiple source domains, generalizing the trained model to an unseen target domain. However, the spurious correlation usually caused by context prior (e.g., background) makes it challenging to get rid of the domain shift. Therefore, it is critical to model the intrinsic causal mechanism. The existing domain generalization methods only attend to disentangle the semantic and context-related features by modeling the causation between input and labels, which totally ignores the unidentifiable but important confounders. In this article, a Causal Disentangled Intervention Model (CDIM) is proposed for the first time, to the best of our knowledge, to construct confounders via causal intervention. Specifically, a generative model is employed to disentangle the semantic and context-related features. The contextual information of each domain from generative model can be considered as a confounder layer, and the center of all context-related features is utilized for fine-grained hierarchical modeling of confounders. Then the semantic and confounding features from each layer are combined to train an unbiased classifier, which exhibits both transferability and robustness across an unknown distribution domain. CDIM is evaluated on three widely recognized benchmark datasets, namely, Digit-DG, PACS, and NICO, through extensive ablation studies. The experimental results clearly demonstrate that the proposed model achieves state-of-the-art performance.

Single domain generalization method based on anti-causal learning for rotating machinery fault diagnosis

Single-domain generalization (SDG) fault diagnosis methods are promising approach because they can diagnose unknown domains by training only one domain. However, there have been fewer studies on fault diagnosis for SDG. Existing studies focus on extending the distribution of source domain, but fail to consider how to estimate domain shifts between domains. Therefore, we propose an SDG fault diagnosis learning paradigm, i.e., simulation-inference-adaptation, which first establishes a pseudo-domain as the target domain to simulate the domain shifts, then reasons the causes of the domain shifts, and finally performs the domain adaptation. In this paradigm, an anti-causal learning approach is proposed, whereby the cause of the domain shifts between the pseudo-domain and the source domain is inferred during training and the learned knowledge is used to analyze the cause of the domain shifts between the target domain and the source domain during testing. Specifically, data augmentation is utilized to perform combinatorial transformation of source data to generate pseudo-domains, perform anti-causal inference to learn to discover causal factors for domain shift between pseudo-domains and source domains, and impose the inferred causality into a weighted domain adaptation network. Finally, the experimental demonstration validated the proposed method's validity and its ability to inference causality.

SAMCF: Adaptive global style alignment and multi-color spaces fusion for joint optic cup and disc segmentation

The optic cup (OC) and optic disc (OD) are two critical structures in retinal fundus images, and their relative positions and sizes are essential for effectively diagnosing eye diseases. With the success of deep learning in computer vision, deep learning-based segmentation models have been widely used for joint optic cup and disc segmentation. However, there are three prominent issues that impact the segmentation performance. First, significant differences among datasets collecting from various institutions, protocols, and devices lead to performance degradation of models. Second, we find that images with only RGB information struggle to counteract the interference caused by brightness variations, affecting color representation capability. Finally, existing methods typically ignored the edge perception, facing the challenges in obtaining clear and smooth edge segmentation results. To address these drawbacks, we propose a novel framework based on Style Alignment and Multi-Color Fusion (SAMCF) for joint OC and OD segmentation. Initially, we introduce a domain generalization method to generate uniformly styled images without damaged image content for mitigating domain shift issues. Next, based on multiple color spaces, we propose a feature extraction and fusion network aiming to handle brightness variation interference and improve color representation capability. Lastly, an edge aware loss is designed to generate fine edge segmentation results. Our experiments conducted on three public datasets, DGS, RIM, and REFUGE, demonstrate that our proposed SAMCF achieves superior performance to existing state-of-the-art methods. Moreover, SAMCF exhibits remarkable generalization ability across multiple retinal fundus image datasets, showcasing its outstanding generality.

AsdinNorm: A Single-Source Domain Generalization Method for the Remaining Useful Life Prediction of Bearings

The remaining useful life (RUL) of bearings is vital for the manipulation and maintenance of industrial machines. The existing domain adaptive methods have achieved major achievements in predicting RUL to tackle the problem of data distribution discrepancy between training and testing sets. However, they are powerless when the target bearing data are not available or unknown for model training. To address this issue, we propose a single-source domain generalization method for RUL prediction of unknown bearings, termed as the adaptive stage division and parallel reversible instance normalization model. First, we develop the instance normalization of the vibration data from bearings to increase data distribution diversity. Then, we propose an adaptive threshold-based degradation point identification method to divide the healthy and degradation stages of the run-to-failure vibration data. Next, the data from degradation stages are selected as training sets to facilitate the RUL prediction of the model. Finally, we combine instance normalization and instance denormalization of the bearing data into a unified GRU-based RUL prediction network for the purpose of leveraging the distribution bias in instance normalization and improving the generalization performance of the model. We use two public datasets to verify the proposed method. The experimental results demonstrate that, in the IEEE PHM Challenge 2012 dataset experiments, the prediction accuracy of our model with the average RMSE value is 1.44, which is 11% superior to that of the suboptimal comparison model (Transformer model). It proves that our model trained on one-bearing data achieves state-of-the-art performance in terms of prediction accuracy on multiple bearings.