Cross-domain Datasets Research Articles

Maximizing RAG efficiency: A comparative analysis of RAG methods

Abstract This paper addresses the optimization of retrieval-augmented generation (RAG) processes by exploring various methodologies, including advanced RAG methods. The research, driven by the need to enhance RAG processes as highlighted by recent studies, involved a grid-search optimization of 23,625 iterations. We evaluated multiple RAG methods across different vectorstores, embedding models, and large language models, using cross-domain datasets and contextual compression filters. The findings emphasize the importance of balancing context quality with similarity-based ranking methods, as well as understanding tradeoffs between similarity scores, token usage, runtime, and hardware utilization. Additionally, contextual compression filters were found to be crucial for efficient hardware utilization and reduced token consumption, despite the evident impacts on similarity scores, which may be acceptable depending on specific use cases and RAG methods.

Generalization Investigation in Heart Disease Prediction: Comparative Analysis of MLP, Random Forest, and SVM with Cross-Domain Datasets

The traditional way of predicting heart disease is usually through the subjective judgment of doctors, which is high subjectivity. Current machine learning methods do not pay enough attention to its universality. The two datasets used in this study are from Kaggle and preprocessed in advance, including standardization by using Z-Score and dimensionality reduction by KPCA. The two datasets were designated as the source domain and the target domain. The source domain dataset was further divided into two subsets with an 80:20 split, where 80% was used for training the model and the remaining 20% served as the test set. The trained model was then applied to the target domain to compare the differences in prediction results. Through the exploration of Multilayer Perceptron (MLP), Random Forest and Support Vector Machine (SVM), for the data set used in this study, when MLP uses a simple structure, the difference in prediction accuracy dropped from 13.8% to 5.07. For Random Forest, by increasing the number of decision trees and decreasing the minimum number of samples required for splitting, the difference is reduced from 15.6% to 9.32%. By modifying the penalty parameter C value of SVM, the difference on different datasets is reduced from 13.39% to 4.46%. This study is one of the few to explore the generalizations of heart disease prediction. The results demonstrate that the generalization performance of the model for heart disease prediction can be significantly enhanced through appropriate modifications to its structure and hyperparameters.

In-domain versus out-of-domain transfer learning for document layout analysis

AbstractData availability is a big concern in the field of document analysis, especially when working on tasks that require a high degree of precision when it comes to the definition of the ground truths on which to train deep learning models. A notable example is represented by the task of document layout analysis in handwritten documents, which requires pixel-precise segmentation maps to highlight the different layout components of each document page. These segmentation maps are typically very time-consuming and require a high degree of domain knowledge to be defined, as they are intrinsically characterized by the content of the text. For this reason in the present work, we explore the effects of different initialization strategies for deep learning models employed for this type of task by relying on both in-domain and cross-domain datasets for their pre-training. To test the employed models we use two publicly available datasets with heterogeneous characteristics both regarding their structure as well as the languages of the contained documents. We show how a combination of cross-domain and in-domain transfer learning approaches leads to the best overall performance of the models, as well as speeding up their convergence process.

Cross-Scene Joint Classification of Multisource Data With Multilevel Domain Adaption Network.

Domain adaption (DA) is a challenging task that integrates knowledge from source domain (SD) to perform data analysis for target domain. Most of the existing DA approaches only focus on single-source-single-target setting. In contrast, multisource (MS) data collaborative utilization has been extensively used in various applications, while how to integrate DA with MS collaboration still faces great challenges. In this article, we propose a multilevel DA network (MDA-NET) for promoting information collaboration and cross-scene (CS) classification based on hyperspectral image (HSI) and light detection and ranging (LiDAR) data. In this framework, modality-related adapters are built, and then a mutual-aid classifier is used to aggregate all the discriminative information captured from different modalities for boosting CS classification performance. Experimental results on two cross-domain datasets show that the proposed method consistently provides better performance than other state-of-the-art DA approaches.

Invariant Content Representation for Generalizable Medical Image Segmentation.

Domain generalization (DG) for medical image segmentation due to privacy preservation prefers learning from a single-source domain and expects good robustness on unseen target domains. To achieve this goal, previous methods mainly use data augmentation to expand the distribution of samples and learn invariant content from them. However, most of these methods commonly perform global augmentation, leading to limited augmented sample diversity. In addition, the style of the augmented image is more scattered than the source domain, which may cause the model to overfit the style of the source domain. To address the above issues, we propose an invariant content representation network (ICRN) to enhance the learning of invariant content and suppress the learning of variability styles. Specifically, we first design a gamma correction-based local style augmentation (LSA) to expand the distribution of samples by augmenting foreground and background styles, respectively. Then, based on the augmented samples, we introduce invariant content learning (ICL) to learn generalizable invariant content from both augmented and source-domain samples. Finally, we design domain-specific batch normalization (DSBN) based style adversarial learning (SAL) to suppress the learning of preferences for source-domain styles. Experimental results show that our proposed method improves by 8.74% and 11.33% in overall dice coefficient (Dice) and reduces 15.88 mm and 3.87 mm in overall average surface distance (ASD) on two publicly available cross-domain datasets, Fundus and Prostate, compared to the state-of-the-art DG methods.The code is available at https://github.com/ZMC-IIIM/ICRN-DG .

Two-stage adversarial learning based unsupervised domain adaptation for retinal OCT segmentation.

Deep learning based optical coherence tomography (OCT) segmentation methods have achieved excellent results, allowing quantitative analysis of large-scale data. However, OCT images are often acquired by different devices or under different imaging protocols, which leads to serious domain shift problem. This in turn results in performance degradation of segmentation models. Aiming at the domain shift problem, we propose a two-stage adversarial learning based network (TSANet) that accomplishes unsupervised cross-domain OCT segmentation. In the first stage, a Fourier transform based approach is adopted to reduce image style differences from the image level. Then, adversarial learning networks, including a segmenter and a discriminator, are designed to achieve inter-domain consistency in the segmentation output. In the second stage, pseudo labels of selected unlabeled target domain training data are used to fine-tune the segmenter, which further improves its generalization capability. The proposed method was tested on cross-domain datasets for choroid or retinoschisis segmentation tasks. For choroid segmentation, the model was trained on 400 images and validated on 100 images from the source domain, and then trained on 1320 unlabeled images and tested on 330 images from target domain I, and also trained on 400 unlabeled images and tested on 200 images from target domain II. For retinoschisis segmentation, the model was trained on 1284 images and validated on 312 images from the source domain, and then trained on 1024 unlabeled images and tested on 200 images from the target domain. The proposed method achieved significantly improved results over that without domain adaptation, with improvement of 8.34%, 55.82% and 3.53% in intersection over union (IoU) respectively for the three test sets. The performance is better than some state-of-the-art domain adaptation methods. The proposed TSANet, with image level adaptation, feature level adaptation and pseudo-label based fine-tuning, achieved excellent cross-domain generalization. This alleviates the burden of obtaining additional manual labels when adapting the deep learning model to new OCT data.

Multiscale Change Detection Domain Adaptation Model Based on Illumination–Reflection Decoupling

In the change detection (CD) task, the substantial variation in feature distributions across different CD datasets significantly limits the reusability of supervised CD models. To alleviate this problem, we propose an illumination–reflection decoupled change detection multi-scale unsupervised domain adaptation model, referred to as IRD-CD-UDA. IRD-CD-UDA maintains its performance on the original dataset (source domain) and improves its performance on unlabeled datasets (target domain) through a novel CD-UDA structure and methodology. IRD-CD-UDA synergizes mid-level global feature marginal distribution domain alignment, classifier layer feature conditional distribution domain alignment, and an easy-to-hard sample selection strategy to increase the generalization performance of CD models on cross-domain datasets. Extensive experiments conducted on the LEVIR, SYSU, and GZ optical remote sensing image datasets demonstrate that the IRD-CD-UDA model effectively mitigates feature distribution discrepancies between source and target CD data, thereby achieving optimal recognition performance on unlabeled target domain datasets.

DomainFeat: Learning Local Features With Domain Adaptation

Accurate and efficient keypoint detection and description is a fundamental step in various computer vision tasks. In this paper, we extract robust descriptors and detect accurate keypoints by learning local Features with Domain adaptation (DomainFeat). Specifically, our Domainfeat includes image-level domain invariance supervision, pixel-level domain consistency supervision, Pixel-Adaptive keypoint Detection(PA-Det), and cross-domain dataset with domain stable point supervision. First, we introduce the image-level domain invariance supervision to make the high-level feature distributions from different domains close by fusing domain-invariant representations in the decoder. Furthermore, to compensate for the inconsistency between descriptors corresponding to the keypoints at the pixel level, we propose the pixel-level domain consistency supervision. Then we present the Pixel-Adaptive keypoint Detection to efficiently detect accurate keypoints, which can improve accuracy by enhancing the local consistency of heatmaps. Finally, we propose an efficient approach to construct data and supervision labels in diverse domains, which can tackle complex application scenarios. With these novel modules and supervision methods, our DomainFeat can make feature detectors more accurate and descriptors more robust. Extensive experiments confirm that Domainfeat achieves state-of-the-art performance on benchmarks such as Aachen-Day-Night localization, HPatches image matching, and the challenging DNIM dataset.

A Cross-Domain Change Detection Network Based on Instance Normalization

Change detection is a crucial task in remote sensing that finds broad application in land resource planning, forest resource monitoring, natural disaster monitoring, and evaluation. In this paper, we propose a change detection model for cross-domain recognition, which we call CrossCDNet. Our model significantly improves the modeling ability of the change detection on one dataset and demonstrates good generalization on another dataset without any additional operations. To achieve this, we employ a Siamese neural network for change detection and design an IBNM (Instance Normalization and Batch Normalization Module) that utilizes instance normalization and batch normalization in order to serve as the encoder backbone in the Siamese neural network. The IBNM extracts feature maps for each layer, and the Siamese neural network fuses the feature maps of the two branches using a unique operation. Finally, a simple MLP decoder is used for end-to-end change detection. We train our model on the LEVIR-CD dataset and achieve competitive performance on the test set. In cross-domain dataset testing, CrossCDNet outperforms all the other compared models. Specifically, our model achieves an F1-score of 91.69% on the LEVIR-CD dataset and an F1-score of 77.09% on the WHU-CD dataset, where the training set was LEVIR-CD.

Plant and Disease Recognition Based on PMF Pipeline Domain Adaptation Method: Using Bark Images as Meta-Dataset.

Efficient image recognition is important in crop and forest management. However, it faces many challenges, such as the large number of plant species and diseases, the variability of plant appearance, and the scarcity of labeled data for training. To address this issue, we modified a SOTA Cross-Domain Few-shot Learning (CDFSL) method based on prototypical networks and attention mechanisms. We employed attention mechanisms to perform feature extraction and prototype generation by focusing on the most relevant parts of the images, then used prototypical networks to learn the prototype of each category and classify new instances. Finally, we demonstrated the effectiveness of the modified CDFSL method on several plant and disease recognition datasets. The results showed that the modified pipeline was able to recognize several cross-domain datasets using generic representations, and achieved up to 96.95% and 94.07% classification accuracy on datasets with the same and different domains, respectively. In addition, we visualized the experimental results, demonstrating the model's stable transfer capability between datasets and the model's high visual correlation with plant and disease biological characteristics. Moreover, by extending the classes of different semantics within the training dataset, our model can be generalized to other domains, which implies broad applicability.

Cross-domain video action recognition via adaptive gradual learning

Video-based Unsupervised Domain Adaptation (UDA) methods concentrate on addressing domain shift and improving the robustness of video models. It can be naturally applied to cross-domain action recognition tasks, but the inherent complexity of videos makes this task more challenging. Though several recent attempts have achieved superior performance in the video UDA field, most of them intend to mitigate domain shift directly from the source domain to the target domain. These methods will cause a performance drop when the domain distribution shift is large. To better realize the domain adaptation, this paper proposes an effective gradual learning framework by constructing multiple auxiliary domains to achieve progressive transfer for cross-domain video action recognition. Specifically, a Dynamic CutMix Mechanism (DCM) is introduced to build the auxiliary domains that mitigate the domain gap caused by object distance and background discrepancies. Furthermore, the Gradual Transfer Strategy (GTS) utilizes these auxiliary domains to realize the cross-domain action classification gradually. Extensive experiments validate the effectiveness of our proposed method, and the experimental results can significantly outperform state-of-the-art methods on multiple benchmark cross-domain datasets.

CGFNet: 3D Convolution Guided and Multi-scale Volume Fusion Network for fast and robust stereo matching

Nowadays, although significant progress has been made by convolutional neural network, it is still difficult to realize accurate and robust stereo matching in real time. In this article, we study how to achieve more accurate and robust disparity estimation based on real-time requirement. For this reason, a Multi-scale Volume Fusion (MVF) module was proposed and embedded to improve the matching accuracy. To achieve real-time performance, an innovative way to use 3D convolution is proposed. The 3D convolution is used during training for guidance and supervision, making the inference lightweight. Based on these two structures, we designed an end-to-end stereo matching method called 3D Convolution Guided and Multi-scale Cost Volume Fusion Network (CGFNet). Experimental results showed that our CGFNet has better generalization performance on cross-domain datasets, which achieves more accurate disparity estimation without additional fine tuning process in challenging regions. On KITTI benchmark, CGFNet reached D1-all=1.98% with substantial improvement among the State-Of-The-Art (SOTA) real-time models and runs a pair of images within 38 ms (26 fps). The results are notable when considering both matching accuracy and real-time performance.

Unsupervised Cross-Domain Rumor Detection with Contrastive Learning and Cross-Attention

Massive rumors usually appear along with breaking news or trending topics, seriously hindering the truth. Existing rumor detection methods are mostly focused on the same domain, thus have poor performance in cross-domain scenarios due to domain shift. In this work, we propose an end-to-end instance-wise and prototype-wise contrastive learning model with cross-attention mechanism for cross-domain rumor detection. The model not only performs cross-domain feature alignment, but also enforces target samples to align with the corresponding prototypes of a given source domain. Since target labels in a target domain are unavailable, we use a clustering-based approach with carefully initialized centers by a batch of source domain samples to produce pseudo labels. Moreover, we use a cross-attention mechanism on a pair of source data and target data with the same labels to learn domain-invariant representations. Because the samples in a domain pair tend to express similar semantic patterns especially on the people’s attitudes (e.g., supporting or denying) towards the same category of rumors, the discrepancy between a pair of source domain and target domain will be decreased. We conduct experiments on four groups of cross-domain datasets and show that our proposed model achieves state-of-the-art performance.

An adaptive growing grid model for a non-stationary environment

The self-organizing map (SOM) represents high-dimensional input samples by a 2-dimensional output topological structure, whereby similar input samples are mapped onto the same output unit or neighboring units on a map for visualization. Although many extended SOM models have been proposed, the need to determine a grid structure of SOM before learning, and the lack of adaptability to rapid changes in input data have not yet been fully overcome. This research proposes an adaptive growing grid (AGG) model, which is a novel neural self-organizing map (SOM), for projecting high-dimensional input samples onto an output grid. Due to the need for a grid structure for visualization, the AGG uses both growing and pruning functions and an adaptive learning process in order to adapt its output grid structure and learning function to constantly and rapidly changing input data in a non-stationary environment. The proposed AGG is tested on four basic data sets and one cross-domain data set. In addition, the t-test is used to test whether the proposed AGG outperforms the benchmark model, the growing grid (GG). Based on three evaluation measures, i.e. average quantization error (AQE), topographic error (TE) and dead unit ratio (DUR), the AGG significantly outperforms the GG in a non-stationary environment.

Handling catastrophic forgetting using cross-domain order in incremental deep learning

In the present era of big data applications, incremental learning has emerged as the most admired area of research where ever-ending tasks from different application domains arrive temporarily, and the learning models majorly focus on newly arriving tasks while forgetting the historically acquired knowledge of old tasks. In this context, when a deep neural network (DNN) performs a sequence of learning tasks, it usually forgets the previously learned tasks. This phenomenon is known as catastrophic forgetting (CF), and the existing state-of-the-art deep learning approaches are unable to handle this issue. Following this, many DNN-based state-of-the-art solutions have been proposed to alleviate the problem that ensures a more natural way of human brain imitation. During training, deep learning algorithms are subjected to a random sequence of diversified datasets (also referred to as cross-domain datasets) that typically belong to multiple domains and have a variety of feature representation, distribution, density, etc. In such cases, CF may become a serious issue because the model tends to forget the previously learned features that affect the overall performance. Because there has been very little research in this area, a more robust incremental deep learning algorithm is proposed in our study to learn desired tasks incrementally. To overcome this situation, proper sequencing of cross-domain datasets is important so that related datasets are kept together to enhance the learnability of the model. Our research proposes a cross-domain order approach with hard attention to the task (HAT) methodology that uses an attention-gated mechanism to efficiently learn the task-specific features. Intense experiments have been performed, and the performance of the proposed approach is compared with the changing cross-domain orders with the baseline models over the following datasets: MNIST, P-MNIST, CIFAR-10, CIFAR-100, and SVHN, etc. The results of the experiments show that learning through cross-domain orders results in better incremental learning.

Roman Urdu Hate Speech Detection Using Transformer-Based Model for Cyber Security Applications.

Social media applications, such as Twitter and Facebook, allow users to communicate and share their thoughts, status updates, opinions, photographs, and videos around the globe. Unfortunately, some people utilize these platforms to disseminate hate speech and abusive language. The growth of hate speech may result in hate crimes, cyber violence, and substantial harm to cyberspace, physical security, and social safety. As a result, hate speech detection is a critical issue for both cyberspace and physical society, necessitating the development of a robust application capable of detecting and combating it in real-time. Hate speech detection is a context-dependent problem that requires context-aware mechanisms for resolution. In this study, we employed a transformer-based model for Roman Urdu hate speech classification due to its ability to capture the text context. In addition, we developed the first Roman Urdu pre-trained BERT model, which we named BERT-RU. For this purpose, we exploited the capabilities of BERT by training it from scratch on the largest Roman Urdu dataset consisting of 173,714 text messages. Traditional and deep learning models were used as baseline models, including LSTM, BiLSTM, BiLSTM + Attention Layer, and CNN. We also investigated the concept of transfer learning by using pre-trained BERT embeddings in conjunction with deep learning models. The performance of each model was evaluated in terms of accuracy, precision, recall, and F-measure. The generalization of each model was evaluated on a cross-domain dataset. The experimental results revealed that the transformer-based model, when directly applied to the classification task of the Roman Urdu hate speech, outperformed traditional machine learning, deep learning models, and pre-trained transformer-based models in terms of accuracy, precision, recall, and F-measure, with scores of 96.70%, 97.25%, 96.74%, and 97.89%, respectively. In addition, the transformer-based model exhibited superior generalization on a cross-domain dataset.

Topic Classification for Political Texts with Pretrained Language Models

AbstractSupervised topic classification requires labeled data. This often becomes a bottleneck as high-quality labeled data are expensive to acquire. To overcome the data scarcity problem, scholars have recently proposed to use cross-domain topic classification to take advantage of preexisting labeled datasets. Cross-domain topic classification only requires limited annotation in the target domain to verify its cross-domain accuracy. In this letter, we propose supervised topic classification with pretrained language models as an alternative. We show that language models fine-tuned with 70% of the small annotated dataset in the target corpus could outperform models trained using large cross-domain datasets by 27% and that models fine-tuned with 10% of the annotated dataset could already outperform the cross-domain classifiers. Our models are competitive in terms of training time and inference time. Researchers interested in supervised learning with limited labeled data should find our results useful. Our code and data are publicly available.1

Importance filtered soft label-based deep adaptation network

Deep adaptation networks exploit domain-invariant features of the source and target domains with some label-induced losses that need ground-truth labels as inputs. As target ground-truth labels are unavailable, existing approaches often utilize target-predicted labels that could be expressed in different forms, to formulate the label-induced losses. Specifically, the form of a hard label classifies a sample into one definite category, but it is overconfident when the quality of predicted labels is poor. In contrast, the form of a soft label gives probability values to each category for a sample, which could mitigate the overconfident problem. However, the soft label would introduce too many small and noisy probability values. Therefore, both label forms may mislead the optimizing process of label-induced losses and produce a negative adaptation. To deal with this challenge, this paper maximizes the nuclear norm of a soft label matrix to filter noisy probability values out while only preserving those important ones, to produce a more reliable form of importance filtered soft label. Then, we utilize the proposed label form to reformulate several prevalent label-induced losses in a probability-weighted manner, so that they could be optimized more correctly during the training process and alleviate the negative adaptation accordingly. Besides, we conduct the proposed scheme in a deep neural network dubbed as importance filtered soft label-based deep adaptation network (IFSL-DAN). Finally, we evaluate the proposed IFSL-DAN with extensive experiments on three cross-domain datasets to demonstrate its effectiveness compared with some state-of-art domain adaptation methods.

Progressive cross-domain knowledge distillation for efficient unsupervised domain adaptive object detection

Unsupervised domain adaptation (UDA) is a technique for relieving domain shifts via transferring relevant domain knowledge from the full-labeled source domain to an unlabeled target domain. While tremendous advances have been witnessed recently, the adoption of deep CNN-based UDA methods in real-world scenarios is still constrained by low-resource computers. Most prior strategies either handle domain shift problems via UDA or compress CNNs using knowledge distillation (KD), we seek to implement the model on constrained-resource devices to learn domain adaptive knowledge without sacrificing accuracy. In this paper, we proposed a three-step Progressive Cross-domain Knowledge Distillation (PCdKD) paradigm for efficient unsupervised adaptive object detection, since directly alleviating the significant discrepancy across domains could result in unstable training procedures and suboptimal performance. First, we apply pixel-level alignment via image-to-image translation to reduce the appearance discrepancy between different domains. Then, a focal multi-domain discriminator is utilized to train the teacher–student peer networks for gradually distilling domain adaptive knowledge in a cooperative manner. Finally, reliable pseudo labels obtained by the adapted teacher detector are further utilized to retrain the teacher–student models. Our proposed method can boost the transferability of the teacher model as well as enhance the student model to meet the demand of real-time applications. Comprehensive experiments on four different cross-domain datasets show that our PCdKD outperforms most existing state-of-the-art approaches.

PCLUDA: A Pseudo-Label Consistency Learning- Based Unsupervised Domain Adaptation Method for Cross-Domain Optical Remote Sensing Image Retrieval

Recent advances in deep learning have dramatically improved the performance of content-based remote sensing image retrieval (CBRSIR) with the same distribution of training set (source domain) and test set (target domain). In fact, their distributions are inconsistent in most cases, which can lead to a dramatic decrease in retrieval performance. Currently, some unsupervised domain adaptation (DA) methods for other remote sensing applications have been proposed to eliminate the inconsistency. However, the current unsupervised DA methods do not make full use of the target domain’s distribution characteristics when delineating its decision boundary. This tends to degrade the cross-domain retrieval performance. In this article, a pseudo-label consistency learning-based unsupervised DA method (PCLUDA) is proposed for cross-domain CBRSIR. Our PCLUDA method minimizes the difference in probability distribution between the target domain and its perturbed output by a pseudo-label self-training and consistency regularization strategy, followed by adjusting the target domain’s decision boundaries to the low-density region. Besides, minimize class confusion (MCC) is introduced to reduce negative transfer caused by large intraclass variance of RSIs. Two cross-domain datasets with 12 cross-domain scenarios are constructed based on six open access datasets to measure DA methods. Experimental results show that our PCLUDA method achieves superior retrieval performances with average retrieval precision improvement by 4.9%–32.3% compared with eight state-of-the-art DA approaches in complex cross-domain scenarios. Furthermore, other experimental results indicate that our PCLUDA can also reach optimal retrieval performances in different kinds of deep learning networks [i.e., vision transformer (ViT) and convolutional neural networks (CNNs)].