Improving the performance of drifted/shifted electronic nose systems by cross-domain transfer using common transfer samples

We investigate the capabilities of transfer learning in the area of structural health monitoring. In particular, we are interested in damage detection for concrete structures. Typical image datasets for such problems are relatively small, calling for the transfer of learned representation from a related large-scale dataset. Past efforts of damage detection using images have mainly considered cross-domain transfer learning approaches using pre-trained ImageNet models that are subsequently fine-tuned for the target task. However, there are rising concerns about the generalizability of ImageNet representations for specific target domains, such as for visual inspection and medical imaging. We, therefore, evaluate a combination of in-domain and cross-domain transfer learning strategies for damage detection in bridges. We perform comprehensive comparisons to study the impact of cross-domain and in-domain transfer, with various initialization strategies, using six publicly available visual inspection datasets. The pre-trained models are also evaluated for their ability to cope with the extremely low-data regime. We show that the combination of cross-domain and in-domain transfer persistently shows superior performance specially with tiny datasets. Likewise, we also provide visual explanations of predictive models to enable algorithmic transparency and provide insights to experts about the intrinsic decision logic of typically black-box deep models.

Considering insufficient data and difficulty of labeling in Synthetic Aperture Radar (SAR) images, we propose a method for SAR ship instance segmentation based on cross-domain transfer learning. Compared with optical images, transfer learning in SAR images faces the difficulties of insufficient data to pre-train and lacking detail features. The proposed method, containing sample transfer module and knowledge transfer module, simulates images from optics to SAR and pre-train the ship detection part of the instance segmentation network with simulation images. In addition, we design a Res-Pyramid network to prevent the deep network from being unable to extract efficient features of SAR images. The method proposed combines the content of the optics and the style of the SAR and incorporates multiscale features in backbone, which improves performance in ship instance segmentation in SAR images. Experiments show that it has achieved 1.3 and 1.1 points higher Average Precision (AP) in detection and segmentation tasks on SAR dataset of HRSID when using cross-domain transfer learning, which has exceeded state-of-the-art methods.

This study aimed to find suitable source domain data in cross-domain transfer learning to extract robust image features. Then, a model was built to preoperatively distinguish lung granulomatous nodules (LGNs) from lung adenocarcinoma (LAC) in solitary pulmonary solid nodules (SPSNs). Data from 841 patients with SPSNs from five centres were collected retrospectively. First, adaptive cross-domain transfer learning was used to construct transfer learning signatures (TLS) under different source domain data and conduct a comparative analysis. The Wasserstein distance was used to assess the similarity between the source domain and target domain data in cross-domain transfer learning. Second, a cross-domain transfer learning radiomics model (TLRM) combining the best performing TLS, clinical factors and subjective CT findings was constructed. Finally, the performance of the model was validated through multicentre validation cohorts. Relative to other source domain data, TLS based on lung whole slide images as source domain data (TLS-LW) had the best performance in all validation cohorts (AUC range: 0.8228-0.8984). Meanwhile, the Wasserstein distance of TLS-LW was 1.7108, which was minimal. Finally, TLS-LW, age, spiculated sign and lobulated shape were used to build the TLRM. In all validation cohorts, The AUC ranges were 0.9074-0.9442. Compared with other models, decision curve analysis and integrated discrimination improvement showed that TLRM had better performance. The TLRM could assist physicians in preoperatively differentiating LGN from LAC in SPSNs. Furthermore, compared with other images, cross-domain transfer learning can extract robust image features when using lung whole slide images as source domain data and has a better effect.

Deep neural networks (DNNs) have achieved remarkable success in various domains; however, their performance often relies heavily on large-scale, high-quality labeled datasets, which are scarce in low-resource environments. Cross-domain transfer learning has emerged as a promising technique for adapting pre-trained models from data-rich source domains to low-resource target domains to address this limitation. This study explores innovative strategies to enhance the performance and applicability of DNNs through cross-domain transfer learning, focusing on challenges such as domain disparity, data scarcity, and computational constraints. We evaluate several transfer learning approaches, including feature-based transfer, parameter fine-tuning, and adversarial domain adaptation, across diverse healthcare, agriculture, and natural language processing applications. Experimental results demonstrate significant improvements in model accuracy and generalization in low-resource environments, with accuracy gains of up to 20% compared to models trained from scratch. Additionally, we analyze the impact of transfer learning on reducing training time and computational requirements, making it a viable solution for resource-constrained settings. Despite its potential, the study highlights critical challenges, including negative transfer, model interpretability, and ethical considerations in domain transfer. Addressing these issues, we propose a framework for selecting optimal source domains and enhancing model robustness through hybrid techniques and unsupervised learning. This research emphasizes the transformative potential of cross-domain transfer learning in bridging the gap between data-rich and low-resource environments, paving the way for more equitable and efficient applications of deep learning technologies worldwide.

We propose a cross-domain, multi-task transfer learning framework to transfer knowledge learned from non-medical images by a deep convolutional neural network (DCNN) to medical image recognition task while improving the generalization by multi-task learning of auxiliary tasks. A first stage cross-domain transfer learning was initiated from ImageNet trained DCNN to mammography trained DCNN. 19,632 regions-of-interest (ROI) from 2,454 mass lesions were collected from two imaging modalities: digitized-screen film mammography (SFM) and full-field digital mammography (DM), and split into training and test sets. In the multi-task transfer learning, the DCNN learned the mass classification task simultaneously from the training set of SFM and DM. The best transfer network for mammography was selected from three transfer networks with different number of convolutional layers frozen. The performance of single-task and multitask transfer learning on an independent SFM test set in terms of the area under the receiver operating characteristic curve (AUC) was 0.78±0.02 and 0.82±0.02, respectively. In the second stage cross-domain transfer learning, a set of 12,680 ROIs from 317 mass lesions on DBT were split into validation and independent test sets. We first studied the data requirements for the first stage mammography trained DCNN by varying the mammography training data from 1% to 100% and evaluated its learning on the DBT validation set in inference mode. We found that the entire available mammography set provided the best generalization. The DBT validation set was then used to train only the last four fully connected layers, resulting in an AUC of 0.90±0.04 on the independent DBT test set.

Automatic biology image classification is essential for biodiversity conservation and ecological study. Recently, due to the record-shattering performance, deep convolutional neural networks (DCNNs) have been used more often in biology image classification. However, training DCNNs requires a large amount of labeled data, which may be difficult to collect for some organisms. This study was carried out to exploit cross-domain transfer learning for DCNNs with limited data. According to the literature, previous studies mainly focus on transferring from ImageNet to a specific domain or transferring between two closely related domains. While this study explores deep transfer learning between species from different domains and analyzes the situation when there is a huge difference between the source domain and the target domain. Inspired by the analysis of previous studies, the effect of biology cross-domain image classification in transfer learning is proposed. In this work, the multiple transfer learning scheme is designed to exploit deep transfer learning on several biology image datasets from different domains. There may be a huge difference between the source domain and the target domain, causing poor performance on transfer learning. To address this problem, multistage transfer learning is proposed by introducing an intermediate domain. The experimental results show the effectiveness of cross-domain transfer learning and the importance of data amount and validate the potential of multistage transfer learning.

A framework for semi-supervised metric transfer learning on manifolds

In recent years, the research on dependency parsing focuses on improving the accuracy of the domain-specific (in-domain) test datasets and has made remarkable progress. However, there are innumerable scenarios in the real world that are not covered by the dataset, namely, the out-of-domain dataset. As a result, parsers that perform well on the in-domain data usually suffer from significant performance degradation on the out-of-domain data. Therefore, to adapt the existing in-domain parsers with high performance to a new domain scenario, cross-domain transfer learning methods are essential to solve the domain problem in parsing. This paper examines two scenarios for cross-domain transfer learning: semi-supervised and unsupervised cross-domain transfer learning. Specifically, we adopt a pre-trained language model BERT for training on the source domain (in-domain) data at the subword level and introduce self-training methods varied from tri-training for these two scenarios. The evaluation results on the NLPCC-2019 shared task and universal dependency parsing task indicate the effectiveness of the adopted approaches on cross-domain transfer learning and show the potential of self-learning to cross-lingual transfer learning.

Transfer learning for multi-objective non-intrusive load monitoring in smart building

In practical bearing fault diagnosis tasks, the available labeled data are often not from the equipment to be diagnosed and cannot cover all manner of working conditions. The adopted data-driven method is required to have a certain degree of cross-domain and cross-working condition transfer learning diagnosis ability. However, limited by the performance of existing transfer learning methods, the potential difference between the source domain and the target domain poses a challenge for the accuracy of transfer diagnosis. In this paper, a cross-working condition data supplement method based on the cycle generative adversarial network (CycleGAN) and a dynamics model is proposed, which can use limited available data to approximate the missing parts of existing data and be used for diagnosis of the target domain. First, we considered the limited experimental data as the target domain, the simulation data corresponding to the working condition as the source domain and used the working condition as the benchmark to constrain the data correspondence between the two datasets. We then used the CycleGAN model to learn the feature mapping from simulation to experiment. Second, based on the working condition of the data to be tested, the corresponding simulation data were input into the trained generator to obtain labeled data with experimental characteristics under the corresponding working conditions, and transferred the dataset as the source domain data to the data to be tested. In the test using self-made simulation and experimental datasets, combined with the transfer learning method based on the probability distribution adaptation, it was shown that the proposed method could effectively improve the diagnostic impact of the single transfer learning method in cross-domain and cross-working conditions when the working condition span was large.

In recent years, the research of dependency parsing focuses on improving the accuracy of in-domain data and has made remarkable progress. However, the real world is different from a single scenario dataset, filled with countless scenarios that are not covered by the dataset, namely, out-of-domain. As a result, parsers that perform well on the in-domain data often suffer significant performance degradation on the out-of-domain data. Therefore, in order to adapt the existing in-domain parsers with substantial performance to the new domain scenario, cross-domain transfer learning techniques are essential to solve the domain problem in parsing. In this paper, we examine two scenarios for cross-domain transfer learning: semi-supervised and unsupervised cross-domain transfer learning. Specifically, we adopt a pretrained language model BERT for training on the source domain (in-domain) data at subword level and introduce two tri-training variant methods for the two scenarios so as to achieve the goal of cross-domain transfer learning. The system based on this paper participated in NLPCC-2019-shared-task on cross-domain dependency parsing and won the first place on the “subtask3-un-open” and “subtask4-semi-open” subtasks, indicating the effectiveness of the approaches adopted.

Pest infestations have always been a major factor affecting tea production. Real-time detection of tea pests using machine vision is a mainstream method in modern agricultural pest control. Currently, there is a notable absence of machine vision devices capable of real-time monitoring for small-sized tea pests in the market, and the scarcity of open-source datasets available for tea pest detection remains a critical limitation. This manuscript proposes a YOLOv8-FasterTea pest detection algorithm based on cross-domain transfer learning, which was successfully deployed in a novel tea pest monitoring device. The proposed method leverages transfer learning from the natural language character domain to the tea pest detection domain, termed cross-domain transfer learning, which is based on the complex and small characteristics shared by natural language characters and tea pests. With sufficient samples in the language character domain, transfer learning can effectively enhance the tiny and complex feature extraction capabilities of deep networks in the pest domain and mitigate the few-shot learning problem in tea pest detection. The information and texture features of small tea pests are more likely to be lost with the layers of a neural network becoming deep. Therefore, the proposed method, YOLOv8-FasterTea, removes the P5 layer and adds a P2 small target detection layer based on the YOLOv8 model. Additionally, the original C2f module is replaced with lighter convolutional modules to reduce the loss of information about small target pests. Finally, this manuscript successfully applies the algorithm to outdoor pest monitoring equipment. Experimental results demonstrate that, on a small sample yellow board pest dataset, the mAP@.5 value of the model increased by approximately 6%, on average, after transfer learning. The YOLOv8-FasterTea model improved the mAP@.5 value by 3.7%, while the model size was reduced by 46.6%.

Automatic Speech Emotion Recognition (SER) plays a vital role in making human–computer interactions more natural and effective. A significant challenge in SER development is the limited availability of diverse emotional speech datasets, which hinders the application of advanced deep learning models. Transfer learning is a machine learning technique that helps address this issue by utilizing knowledge from pre-trained models to improve performance on a new task in a target domain, even with limited data. This study investigates the use of transfer learning from various pre-trained networks, including speaker embedding models such as d-vector, x-vector, and r-vector, and image classification models like AlexNet, GoogLeNet, SqueezeNet, ResNet-18, and ResNet-50. We also propose enhanced versions of the x-vector and r-vector models incorporating Multi-Head Attention Pooling and Angular Margin Softmax, alongside other architectural improvements. Additionally, reverberation from the Room Impulse Response datasets was added to the speech utterances to diversify and augment the available data. Notably, the enhanced r-vector model achieved classification accuracies of 74.05% Unweighted Accuracy (UA) and 73.68% Weighted Accuracy (WA) on the IEMOCAP dataset, and 80.25% UA and 79.81% WA on the CREMA-D dataset, outperforming the existing state-of-the-art methods. This study shows that using cross-domain transfer learning is beneficial for low-resource emotion recognition. The enhanced models developed in other domains (for non-emotional tasks) can further improve the accuracy of SER.

Many existing techniques for recognizing sports posture depend heavily on large, labeled datasets, which limits their performance in scenarios with limited data, such as few-shot learning in specific sports. To address this challenge, we introduce a novel framework that integrates transfer learning, LSTM (Long Short-Term Memory) and GAN (Generative Adversarial Network). The transfer learning component extracts common features from large-scale data in the source domain, while the LSTM module captures temporal dependencies crucial for posture recognition. Simultaneously, the GAN component generates synthetic data to supplement the sparse dataset in the target domain, improving the overall model effectiveness. Experimental results on the Human3.6M and Kinetics-400 datasets show that the proposed TGAN–LSTM model outperforms others in key evaluation metrics, achieving an accuracy of 85.2%, mAP of 79.8% and AUC of 87.3% on Human3.6M, and an accuracy of 80.7% with an F1-score of 78.5% on Kinetics-400. In comparison to other baseline models, TGAN–LSTM demonstrates remarkable performance under few-shot learning conditions. This approach not only offers a solution for few-shot sports posture recognition but also contributes valuable insights to research in cross-domain transfer learning and few-shot learning, with wide practical applications.

Online news media and social media are popular domains for people to acquire real-world event knowledge. In this work, the problem of multi-domain and multi-modality event detection (MMED) is elaborated. We wish to organize the multi-modality data from multiple domains based on real-world events. To this end, a cross-domain and cross-modality transfer learning (CDM) model is proposed. The CDM model aligns the data by exploiting a dictionary-based alignment strategy, and identifies the event labels of the data samples based on the class-specific reconstruction residual. Extensive experiments conducted on real-world data demonstrate the effectiveness of the proposed models. In particular, a benchmark dataset, denoted as MMED100, is released, which can hopefully be used to promote the research on this topic and advance related applications.