Transfer Learning-based Method Research Articles

Towards the generalization of time series classification: A feature-level style transfer and multi-source transfer learning perspective

Transfer learning-based methods hold promise for enhancing classification task performance. However, a transfer learning mechanism for hard-to-classify time series classification tasks caused by limited samples in training set is still to be accomplished. Drawing inspiration from domain adaptation and style transfer methods in computer vision, we aim to address sample scarcity and enhance classifier generalization by leveraging abundant unrelated time series datasets. In this paper, a transfer learning mechanism called feature-level style transfer, where the feature alignment is conducted in a domain adaptation-inspired manner, has been proposed to target initially hard-to-classify time series tasks with limited training samples. The proposed mechanism eliminates constraints on source dataset selection, enabling the utilization of weakly-related or unrelated datasets to enhance classifier performance. Furthermore, a voting mechanism has been formulated to achieve more accurate predictions by taking transferable information learnt from different source domains into comprehensive consideration.

Transfer Learning-Based Structural Damage Identification for Building Structures with Limited Measurement Data

Structural damage detection is crucial for ensuring the safety of civil building structures in operational environments. Recently, deep learning-based methods have gained increasing attention from engineers and researchers. The performance of conventional deep learning methods for structural damage detection relies on a large number of labeled training datasets. However, it is difficult or/and impossible to obtain sufficient datasets to cover various damage scenarios for in-service structures. A little research has been conducted to identify both the damage severity and location with limited labeled measurement data. A novel transfer learning-based method for structural damage identification with limited measurements has been proposed utilizing frequency response functions (FRFs) as the input. The real structure is regarded as the target domain and its numerical model is as the source domain. The samples for various damage scenarios are generated using the numerical model, and a designed deep convolutional neural network (CNN) is pre-trained. The knowledge of the pre-trained network is transferred to identify the damage location and severity of the real structure using limited measurement data. Numerical and experimental studies have been conducted on a three-story building structure to verify the performance of the proposed method. To understand transfer learning and model interpretability, the t-SNE feature visualization is adopted to show the feature distribution changes during transfer learning. Numerical and experimental results show that the proposed approach outperforms conventional CNN models, and it is effective and accurate in identifying structural damage location and severity in real structures with limited measurement data.

Transfer learning-based hybrid deep learning method for gas-bearing distribution prediction with insufficient training samples and uncertainty analysis

The availability of sample data, required for seismicity-based reservoir prediction using deep learning (DL), is often limited by well availability. To alleviate this problem, this study proposes a hybrid DL method combining unsupervised learning, transfer learning (TL), and intelligent optimization algorithms to predict gas-bearing distributions. First, three unsupervised learning methods were used to optimize multicomponent seismic attribute data during preprocessing and improve the sample data quality. Subsequently, a convolutional neural network (CNN) model was pre-trained using synthetic data, and the adaptive mutation particle swarm optimization (AMPSO) algorithm was used to optimize the hyperparameters of the CNN model during training to obtain the AMPSO-CNN model. Finally, the partial convolutional layer parameters of the pre-trained network were transferred, and the remaining parameters were adjusted using real data samples to obtain the TL-AMPSO-CNN model. The trained model was used for gas-bearing distribution prediction, and its stability was assessed using uncertainty analysis. The results confirmed that the hybrid DL model performed efficiently in gas-bearing prediction compared to other individual models. Thus, for datasets with insufficient training samples, the proposed scheme can be applied as a feasible alternative to predict gas-bearing distribution.

A few-shot based phase-batch multi-layer domain adaptation pattern recognition method

Deep transfer learning has been widely applied in the field of intelligent fault diagnosis. However, existing deep transfer learning-based diagnostic methods struggle to train reliable diagnostic models when there is a lack of data and significant distribution differences between the two domains. To address this issue, a few-shot based phase-batch multi-layer domain adaptation pattern recognition method is proposed. This method simultaneously measures the feature distribution differences of both the fully connected layers and the classification layers, thus better correcting the data domain bias. Additionally, a phase-batch training strategy and pseudo-label learning are employed to improve the convergence speed and stability of the training process. The proposed method is validated on two public datasets, Jiang Nan and Paderborn University, as well as a dataset obtained through independent experiments. It is compared with traditional feature-based transfer learning methods, the results show that the proposed method achieves higher diagnostic accuracy, faster convergence, and greater stability. Furthermore, its superior diagnostic performance in the few-shot scenario is demonstrated through experiments on a self-collected dataset.

Study on transfer learning-based cross-spectral speckle image reconstruction method

Abstract In recent years, convolutional neural networks (CNNs) have been successfully applied to reconstruct images from speckle patterns generated as objects pass through scattering media. To achieve this objective, a large amount of data is collected for training the CNN. However, in certain cases, the characteristics of light passing through the scattering medium may vary. In such situations, it is necessary to collect a substantial amount of new data to re-train the CNN and achieve image reconstruction. To address this challenge, transfer learning techniques are introduced in this study. Specifically, we propose a novel Residual U-Net Generative Adversarial Network, denoted as ResU-GAN. The network is initially pre-trained using a large amount of data collected from either visible or non-visible light sources, and subsequently fine-tuned using a small amount of data collected from non-visible or visible light sources. Experimental results demonstrate the outstanding reconstruction performance of the ResU-GAN network. Furthermore, by combining transfer learning techniques, the network enables the reconstruction of speckle images across different datasets. The findings presented in this paper provide a more generalized approach for utilizing CNNs in cross-spectral speckle imaging.

A novel decomposition and hybrid transfer learning-based method for multi-step cutterhead torque prediction of shield machine

As a core operating parameter of shield machine, the high-precision multi-step prediction of cutterhead torque can effectively help the driver adjust the operating parameters in advance to ensure the safe operation of shield machine. Due to data distribution discrepancy, it is challenging to accurately predict the cutterhead torque under different geological conditions. Currently, the existing methods cannot precisely predict cutterhead torque in cross domain. In order to accurately predict the cutterhead torque under different geological environments, we propose a novel decomposition and hybrid transfer learning-based method for multi-step cutterhead torque transfer prediction. First, original cutterhead torque signal is decomposed into some sub-signals by variational mode decomposition (VMD), which can effectively reduce the complexity. Then hybrid transfer learning-based method (HTLM) is established to predict each sub-signal and add the prediction results to realize multi-step cutterhead torque prediction in cross domain. HTLM has three network branches. The first one is that multi-layer GRU network is used to extract time-varying characteristics of source domain data. The second is that the automatic encoder is used to extract the deep common features between source domain and target domain. The third is the adversarial network and the maximum mean discrepancy (MMD) metric for domain adaptation. Then, the extracted features are integrated to predict the cutterhead torque in cross domain. To verify the proposed VMD-HTLM, several transfer prediction experiments were carried out. From the first step to fifth step transfer prediction, the proposed method achieved higher prediction accuracy than existing methods. Especially in the fifth step, the accuracy of proposed method is on average 7.09 % higher than existing methods, the RMSE and the MAE decreased on average 40.00 % and 39.92 %, respectively. Furthermore, the average accuracy of proposed method is 93.79 %, 93.58 %, 93.55 %, 92.25 % and 91.79 % from first step to fifth step, respectively. Hence, the proposed VMD-HTLM can accurately predict the cutterhead torque under different geological conditions.

Efficient-gastro: optimized EfficientNet model for the detection of gastrointestinal disorders using transfer learning and wireless capsule endoscopy images.

Gastrointestinal diseases cause around two million deaths globally. Wireless capsule endoscopy is a recent advancement in medical imaging, but manual diagnosis is challenging due to the large number of images generated. This has led to research into computer-assisted methodologies for diagnosing these images. Endoscopy produces thousands of frames for each patient, making manual examination difficult, laborious, and error-prone. An automated approach is essential to speed up the diagnosis process, reduce costs, and potentially save lives. This study proposes transfer learning-based efficient deep learning methods for detecting gastrointestinal disorders from multiple modalities, aiming to detect gastrointestinal diseases with superior accuracy and reduce the efforts and costs of medical experts. The Kvasir eight-class dataset was used for the experiment, where endoscopic images were preprocessed and enriched with augmentation techniques. An EfficientNet model was optimized via transfer learning and fine tuning, and the model was compared to the most widely used pre-trained deep learning models. The model's efficacy was tested on another independent endoscopic dataset to prove its robustness and reliability.

A Transfer Learning-Based Method for Cyber-Attack Tolerance in Distributed Control of Microgrids

A Transfer Learning-Based Method for Cyber-Attack Tolerance in Distributed Control of Microgrids

A Transfer Learning-based Method for the Daily Electricity Consumption Forecasting of Large Industrial Users after Business Expansion

Background: With the rapid development of industry, the expansion capacity and frequency of large industrial users continue to increase. However, the traditional static prediction model is difficult to accurately predict the daily electricity consumption of industrial expansion, which is not conducive to the safe and stable operation of the power grid. Objective: In response to the above problems, this paper proposes a transfer learning-based method for the daily electricity consumption forecasting of large industrial users after business expansion. Methods: Firstly, a dynamic training framework for the prediction model of transfer learning is established, so that the prediction model can dynamically adapt to the capacity change brought about by the expansion of multi-user business. Then, a neural network for predicting daily electricity consumption of industrial users based on multi-resolution time series attention is established, which can deeply mine the characteristics of electricity sequence. Finally, a deep learning model parameter migration and adjustment method considering business expansion is proposed, which can realize efficient migration of prediction models. Results: The effectiveness of the proposed method is demonstrated by comparing it with state-ofthe- art electricity forecasting based on two-year historical data of a specific region. Conclusion: The proposed method is compared with state-of-the-art power forecasting techniques through the validation of local historical data. The obtained results demonstrate the effectiveness of the proposed method.

Prediction method for formation pore pressure based on transfer learning

Pore pressure prediction of drilled wells is the recognition of formation information using logging and other actual drilling data after drilling is completed, which is of great significance in improving the mastery of regional formation information and reducing the potential risks of the engineering program of wells to be drilled. Traditional formation pressure prediction methods rely on manual experience and are difficult to cope with complex formations. Existing machine learning-based methods have high accuracy in the prediction of formation pressure for wells with similar data composition in the same block or logging, but have poor accuracy in the prediction of wells with large differences in data from different blocks or logging. In this paper, a transfer learning algorithm is introduced to modify the existing LSTM-based formation pore pressure prediction model and establish a transfer learning-based formation pore pressure prediction method. In addition, we evaluated the prediction performance of machine learning algorithms and transfer learning algorithms under the same/different blocks and the same/different logging data conditions. The results show that, compared with the existing machine learning algorithms, for the same block and the same logging data conditions, transfer learning can improve the prediction accuracy by 0.15%; for the same block and different logging data conditions, transfer learning can improve the prediction accuracy by 0.33%; for the different block and the same logging data conditions, transfer learning can improve the prediction accuracy by 0.19%; for the different block and different logging data conditions, transfer learning can improve the prediction accuracy by 0.11%. Overall, the addition of the transfer learning method can improve the accuracy and generalization ability of the formation pore pressure prediction model based on machine learning to a larger extent.

Identifying optical microscope images of CVD-grown two-dimensional MoS2 by convolutional neural networks and transfer learning.

In Complementary Metal-Oxide Semiconductor (CMOS) technology, scaling down has been a key strategy to improve chip performance and reduce power losses. However, challenges such as sub-threshold leakage and gate leakage, resulting from short-channel effects, contribute to an increase in distributed static power. Two-dimensional transition metal dichalcogenides (2D TMDs) emerge as potential solutions, serving as channel materials with steep sub-threshold swings and lower power consumption. However, the production and development of these 2-dimensional materials require some time-consuming tasks. In order to employ them in different fields, including chip technology, it is crucial to ensure that their production meets the required standards of quality and uniformity; in this context, deep learning techniques show significant potential. This research introduces a transfer learning-based deep convolutional neural network (CNN) to classify chemical vapor deposition (CVD) grown molybdenum disulfide (MoS2) flakes based on their uniformity or the occurrence of defects affecting electronic properties. Acquiring and labeling a sufficient number of microscope images for CNN training may not be realistic. To address this challenge, artificial images were generated using Fresnel equations to pre-train the CNN. Subsequently, accuracy was improved through fine-tuning with a limited set of real images. The proposed transfer learning-based CNN method significantly improved all measurement metrics with respect to the ordinary CNNs. The initial CNN, trained with limited data and without transfer learning, achieved 68% average accuracy for binary classification. Through transfer learning and artificial images, the same CNN achieved 85% average accuracy, demonstrating an average increase of approximately 17%. While this study specifically focuses on MoS2 structures, the same methodology can be extended to other 2-dimensional materials by simply incorporating their specific parameters when generating artificial images.

A Novel Transfer Learning-Based Cell SOC Online Estimation Method for a Battery Pack in Complex Application Conditions

The crucial function of a battery management system is to estimate the state of charge (SOC). However, complex application conditions like different temperatures, aging, and inconsistency between cells would cause a distribution difference between data domains and lead to a significant estimation error in the established SOC model. Transfer learning's domain adaptation offers a way to lessen the disparity in distribution across the data domains, and yet it also requires that the target space be the same across domains, which is challenging to do with the SOC online estimate. This article proposes a SOC differential processing method as well as designs a combined transfer learning method. As experiments of actual battery pack provided by China First Automobile Work-based confirmed, the proposed method can obtain precise and robust SOC estimation results with a mean absolute percentage error of less than 1.5%.

A New Framework Based on Supervised Joint Distribution Adaptation for Bearing Fault Diagnosis across Diverse Working Conditions

To address the degradation of diagnostic performance due to data distribution differences and the scarcity of labeled fault data, this study has focused on transfer learning-based cross-domain fault diagnosis, which attracts considerable attention. However, deep transfer learning-based methods often present a challenge due to their time-consuming and costly nature, particularly in tuning hyperparameters. For this issue, on the basis of classical features-based transfer learning method, this study introduces a new framework for bearing fault diagnosis based on supervised joint distribution adaptation and feature refinement. It first utilizes ensemble empirical mode decomposition to process raw signals, and statistical features extraction is implemented. Then, a new feature refinement module is designed to refine domain adaptation features from high-dimensional feature set by evaluating the fault distinguishability and working-condition invariance of feature data. Next, it proposes a supervised joint distribution adaptation method to conduct improved joint distribution alignment that preserves neighborhood relationships within a manifold subspace. Finally, an adaptive classifier is trained to predict fault labels of feature data across varying working conditions. To prove the cross-domain fault diagnosis performance and superiority of the proposed methods, two bearing datasets are applied for experiments, and the experimental results verify that the model built by the proposed framework can achieve desirable diagnosis performance under different working conditions and that it apparently outperforms comparative models.

A Transfer Learning-Based Method for Personalized State of Health Estimation of Lithium-Ion Batteries.

State of health (SOH) estimation of lithium-ion batteries (LIBs) is of critical importance for battery management systems (BMSs) of electronic devices. An accurate SOH estimation is still a challenging problem limited by diverse usage conditions between training and testing LIBs. To tackle this problem, this article proposes a transfer learning-based method for personalized SOH estimation of a new battery. More specifically, a convolutional neural network (CNN) combined with an improved domain adaptation method is used to construct an SOH estimation model, where the CNN is used to automatically extract features from raw charging voltage trajectories, while the domain adaptation method named maximum mean discrepancy (MMD) is adopted to reduce the distribution difference between training and testing battery data. This article extends MMD from classification tasks to regression tasks, which can therefore be used for SOH estimation. Three different datasets with different charging policies, discharging policies, and ambient temperatures are used to validate the effectiveness and generalizability of the proposed method. The superiority of the proposed SOH estimation method is demonstrated through the comparison with direct model training using state-of-the-art machine learning methods and several other domain adaptation approaches. The results show that the proposed transfer learning-based method has wide generalizability as well as a positive precision improvement.

A Cross-Scale Transformer and Triple-View Attention Based Domain-Rectified Transfer Learning for EEG Classification in RSVP Tasks.

Rapid serial visual presentation (RSVP)-based brain-computer interface (BCI) is a promising target detection technique by using electroencephalogram (EEG) signals. However, existing deep learning approaches seldom considered dependencies of multi-scale temporal features and discriminative multi-view spectral features simultaneously, which limits the representation learning ability of the model and undermine the EEG classification performance. In addition, recent transfer learning-based methods generally failed to obtain transferable cross-subject invariant representations and commonly ignore the individual-specific information, leading to the poor cross-subject transfer performance. In response to these limitations, we propose a cross-scale Transformer and triple-view attention based domain-rectified transfer learning (CST-TVA-DRTL) for the RSVP classification. Specially, we first develop a cross-scale Transformer (CST) to extract multi-scale temporal features and exploit the dependencies of different scales features. Then, a triple-view attention (TVA) is designed to capture spectral features from triple views of multi-channel time-frequency images. Finally, a domain-rectified transfer learning (DRTL) framework is proposed to simultaneously obtain transferable domain-invariant representations and untransferable domain-specific representations, then utilize domain-specific information to rectify domain-invariant representations to adapt to target data. Experimental results on two public RSVP datasets suggests that our CST-TVA-DRTL outperforms the state-of-the-art methods in the RSVP classification task. The source code of our model is publicly available in https://github.com/ljbuaa/CST_TVA_DRTL.

STFuse: Infrared and Visible Image Fusion via Semisupervised Transfer Learning.

Infrared and visible image fusion (IVIF) aims to obtain an image that contains complementary information about the source images. However, it is challenging to define complementary information between source images in the lack of ground truth and without borrowing prior knowledge. Therefore, we propose a semisupervised transfer learning-based method for IVIF, termed STFuse, which aims to transfer knowledge from an informative source domain to a target domain, thus breaking the above limitations. The critical aspect of our method is to borrow supervised knowledge from the multifocus image fusion (MFIF) task and to filter out task-specific attribute knowledge by using a guidance loss Lg , which motivates its cross-task use in IVIF tasks. Using this cross-task knowledge effectively alleviates the limitation of the lack of ground truth on fusion performance, and the complementary expression ability under the constraint of supervised knowledge is more instructive than prior knowledge. Moreover, we designed a cross-feature enhancement module (CEM) that utilizes self-attention and mutual-attention features to guide each branch to refine features and then facilitate the integration of cross-modal complementary features. Extensive experiments demonstrate that our method has good advantages in terms of visual quality and statistical metrics, as well as the docking of high-level vision tasks, compared with other state-of-the-art methods.

Understanding Discrepancy of Power System Dynamic Security Assessment with Unknown Faults: A Reliable Transfer Learning-based Method

Understanding Discrepancy of Power System Dynamic Security Assessment with Unknown Faults: A Reliable Transfer Learning-based Method

Transfer learning-based encoder-decoder model with visual explanations for infrastructure crack segmentation: New open database and comprehensive evaluation

Contemporary demands necessitate the swift and accurate detection of cracks in critical infrastructures, including tunnels and pavements. This study proposed a transfer learning-based encoder-decoder method with visual explanations for infrastructure crack segmentation. Firstly, a vast dataset containing 7089 images was developed, comprising diverse conditions—simple and complex crack patterns as well as clean and rough backgrounds. Secondly, leveraging transfer learning, an encoder-decoder model with visual explanations was formulated, utilizing varied pre-trained convolutional neural network (CNN) as the encoder. Visual explanations were achieved through gradient-weighted class activation mapping (Grad-CAM) to interpret the CNN segmentation model. Thirdly, accuracy, complexity (computation and model), and memory usage assessed CNN feasibility in practical engineering. Model performance was gauged via prediction and visual explanation. The investigation encompassed hyperparameters, data augmentation, deep learning from scratch vs. transfer learning, segmentation model architectures, segmentation model encoders, and encoder pre-training strategies. Results underscored transfer learning's potency in enhancing CNN accuracy for crack segmentation, surpassing deep learning from scratch. Notably, encoder classification accuracy bore no significant correlation with CNN segmentation accuracy. Among all tested models, UNet-EfficientNet_B7 excelled in crack segmentation, harmonizing accuracy, complexity, memory usage, prediction, and visual explanation.

Displacement Values Calculation Method for Ship Multi-Support Shafting Based on Transfer Learning

Deviations between the design and actual shafting occur due to limitations in ship construction accuracy. Consequently, accurately obtaining the relationship between the actual shafting load and displacement relationship based on the design shafting becomes challenging, leading to inaccurate solutions for bearing displacement values and low alignment efficiency. In this research article, to address the issue of incomplete actual shafting data, a transfer learning-based method is proposed for accurate calculation of bearing displacement values. By combining simulated data from the design shafting with measured data generated during the adjustment process of the actual shafting, higher accuracy can be achieved in calculating bearing displacement values. This research utilizes a certain shafting as an example to carry out the application of the bearing displacement value calculation method. The results show that even under the action of shafting deviation, the actual shafting load and displacement relationship model can become more and more accurate with the shafting adjustment process, and the accuracy of bearing displacement values calculation becomes higher and higher. This method contributes to obtaining precise shafting adjustment schemes, thereby enhancing alignment quality and efficiency of ship shafting.

Optimized Transfer Learning Based Short-term Electrical Load Interval Prediction

Electrical load forecasting is an essential foundation for power reliable and economical operation of the power grid. Most forecasting models regard the prediction results as deterministic variables, which ignores the randomness and volatility of the power load. At the same time, insufficient historical load data often lead to undertrained models, which affects the accuracy of capturing uncertain information. Therefore, we proposed an optimized transfer learning-based method for short-term load-interval prediction. A deep learning quantile regression model would be constructed by source domain data in the method, and the weights of the source model would be optimized to avoid negative transfer. Then, the target model is constructed by parameter transfer based on key layers and is tuned with hyperparameters by target domain data. From the experimental discussion, it is known that the model with an optimized transfer learning strategy can accurately quantify the fluctuation range of future power load.