Multi-source Transfer Learning Research Articles

Multisource Domain Adaption for Health Degradation Monitoring of Lithium-Ion Batteries

Battery health monitoring is crucial to reduce unexpected maintenance and realize maximum productivity. However, batteries generally work under dynamic and complex working conditions (i.e., different domains) in industrial applications. The distribution difference exists not only between source and target domain but also between multisource domains, it is a challenging problem to monitor battery health deterioration under multiple working conditions. This study proposes a novel transfer learning-based method, multisource domain adaption network (MSDAN), to address the problem of battery health degradation monitoring under different working conditions. First, a battery health degradation monitoring method is proposed based on multisource transfer learning. The knowledge learned from multiple operational conditions is transferred to the target condition. Second, multisource domain adaption is proposed to address the problem of distribution discrepancy of different domains, where unsupervised feature alignment metric and maximum mean discrepancy (MMD) are utilized to evaluate the difference between target domain and each source domain. Multisource adversarial learning is adopted to guide feature generator to provide domain-invariant features of different domains. Third, a feature generator is proposed to extract common and domain-specific features of different domains. The experimental results illustrate that MSDAN has a better performance of battery health prognostics than other single-source domain adaption methods.

Water Quality Prediction Method Based on Multi-Source Transfer Learning for Water Environmental IoT System

Water environmental Internet of Things (IoT) system, which is composed of multiple monitoring points equipped with various water quality IoT devices, provides the possibility for accurate water quality prediction. In the same water area, water flows and exchanges between multiple monitoring points, resulting in an adjacency effect in the water quality information. However, traditional water quality prediction methods only use the water quality information of one monitoring point, ignoring the information of nearby monitoring points. In this paper, we propose a water quality prediction method based on multi-source transfer learning for a water environmental IoT system, in order to effectively use the water quality information of nearby monitoring points to improve the prediction accuracy. First, a water quality prediction framework based on multi-source transfer learning is constructed. Specifically, the common features in water quality samples of multiple nearby monitoring points and target monitoring points are extracted and then aligned. According to the aligned features of water quality samples, the water quality prediction models based on an echo state network at multiple nearby monitoring points are established with distributed computing, and then the prediction results of distributed water quality prediction models are integrated. Second, the prediction parameters of multi-source transfer learning are optimized. Specifically, the back propagates population deviation based on multiple iterations, reducing the feature alignment bias and the model alignment bias to improve the prediction accuracy. Finally, the proposed method is applied in the actual water quality dataset of Hong Kong. The experimental results demonstrate that the proposed method can make full use of the water quality information of multiple nearby monitoring points to train several water quality prediction models and reduce the prediction bias.

A general multi-source ensemble transfer learning framework integrate of LSTM-DANN and similarity metric for building energy prediction

Transfer learning can improve building energy prediction performance by utilizing the knowledge learned from source domain. However, most studies focus on the single-source transfer learning and may lead to model performance degradation when there exists large domain shift between the single source domain and target domain. To address this issue, this study proposes a multi-source ensemble transfer learning (Multi-LSTM-DANN) framework integrate of LSTM-DANN neural network and similarity metric, which can enhance the prediction performance of target building power consumption by using multi-source building data (domain). LSTM-DANN is first used to extract the domain invariant features between each pair of source domain and target domain. Then maximum mean discrepancy (MMD) is applied to metric the distance between each pair of the extracted domain invariant features. Finally, the reciprocal of MMD is used as similarity metric index to calculate the regression weight and prediction value of the proposed Multi-LSTM-DANN model. Experiments with different number of source domains are conducted to demonstrate the effectiveness of the proposed Multi-LSTM-DANN framework. Results demonstrate that most multi-source transfer learning models can enhance the prediction performance of the target building power consumption compared to the corresponding single-source transfer learning models. The proposed Multi-LSTM-DANN framework can provide guiding significance for the application of multi-source building data in the future.

Classification of motor imagery using multisource jointtransfer learning.

As an important way for human-computer interaction, the motor imagery brain-computer interface (MI-BCI) can decode personal motor intention directly by analyzing electroencephalogram (EEG) signals. However, a large amount of labeled data has to be collected for each new subject since EEG patterns vary between individuals. The long calibration phase severely limits the further development of MI-BCI. To tackle this problem, multi-source jointdomain adaption (MJDA) and multi-source jointRiemannian adaption (MJRA) algorithms are proposed in this paper. Both methods aim to transfer knowledge from other subjects to the current subject who has only a small amount of labeled data. First, the common spatial pattern with Euclidean alignment is used to select source subjects who have similar spatial patterns to the target subject. Second, the covariance matrices of EEG trials are aligned in Riemannian space by removing subject-specific baselines. These two steps are shared by MJDA and MJRA. In the last step, MJDA attempts to minimize the feature distribution mismatch in the Riemannian tangent space, while MJRA attempts to find an adaptive Riemannian classifier. Finally, the proposed methods are validated on two datasets: BCI Competition IV 2a and online event-related desynchronization (ERD)-BCI. The experimental results demonstrate that both MJDA and MJRA outperform the state-of-the-art approaches. The MJDA provides a new idea for the offline analysis of MI-BCI, while MJRA could make a big difference to the online calibration of MI-BCI.

Multi-source sintering transfer learning in small dataset sintering prediction scenario

Multi-source sintering transfer learning in small dataset sintering prediction scenario

Multi-source transfer learning network to complement knowledge for intelligent diagnosis of machines with unseen faults

Most of the current successes of deep transfer learning-based fault diagnosis require two assumptions: 1) the health state set of source machines should overlap that of target machines; 2) the number of target machine samples is balanced across health states. However, such assumptions are unrealistic in engineering scenarios, where target machines suffer from fault types that are not seen in source machines and the target machines are mostly in a healthy state with only occasional faults. As a result, the diagnostic knowledge from source machines may not cover all fault types of target machines nor address imbalanced target samples. Therefore, we propose a framework, called a multi-source transfer learning network (MSTLN), to aggregate and transfer diagnostic knowledge from multiple source machines by combining multiple partial distribution adaptation sub-networks (PDA-Subnets) and a multi-source diagnostic knowledge fusion module. The former weights target samples by counter-balancing factors to jointly adapt partial distributions of source and target pairs, and the latter releases negative effects due to discrepancy among multiple source machines and further fuses diagnostic decisions output from multiple PDA-Subnets. Two case studies demonstrate that MSTLN can reduce the misdiagnosis rate and obtain better transfer performance for imbalanced target samples than other conventional methods.

Integrating Multi-Source Transfer Learning, Active Learning and Metric Learning paradigms for Time Series Prediction

Traditional Time Series Prediction (TSP) algorithms assume that the training and testing data follow the same distribution and a large amount of data can be obtained. However, the time-varying nature of time series data is a very difficult problem, which will lead to the inconsistent distribution between new data and old data. In TSP problem scenario, how to transfer knowledge in a relatively long-time span effectively is a serious challenge. To address this issue, in this work, a novel Multi-Source Active Metric Transfer Learning (MS-AMTL) algorithm, integrating Multi-Source Transfer Learning, Active Learning, and Metric Learning paradigms, is designed to make the most of, instead of throw away directly, the long-ago data, with effect, along with its precise mathematic derivation. The setting of multi-source domains can make the dependability of domains and the similarity of sources be fully exerted, reducing the impact of negative transfer. The Active Learning (AL) module in MS-AMTL is designed to select the most representative instances, so that dependability of the domains can be ensured and information redundancy can be reduced. The Metric Learning (ML) module in MS-AMTL is developed to measure the similarity between instances and then the similarity of sources, more effectively. Extensive experiments on one artificial dataset, three real-world datasets and four financial datasets proved the superiority of our proposed algorithm in transfer learning and prediction performance in comparison with several other state-of-the-art algorithms.

MHCPDP: multi-source heterogeneous cross-project defect prediction via multi-source transfer learning and autoencoder

Heterogeneous cross-project defect prediction (HCPDP) is aimed at building a defect prediction model for the target project by reusing datasets from source projects, where the source project datasets and target project dataset have different features. Most existing HCPDP methods only remove redundant or unrelated features without exploring the underlying features of cross-project datasets. Additionally, when the transfer learning method is used in HCPDP, these methods ignore the negative effect of transfer learning. In this paper, we propose a novel HCPDP method called multi-source heterogeneous cross-project defect prediction (MHCPDP). To reduce the gap between the target datasets and the source datasets, MHCPDP uses the autoencoder to extract the intermediate features from the original datasets instead of simply removing redundant and unrelated features and adopts a modified autoencoder algorithm to make instance selection for eliminating irrelevant instances from the source domain datasets. Furthermore, by incorporating multiple source projects to increase the number of source datasets, MHCPDP develops a multi-source transfer learning algorithm to reduce the impact of negative transfers and upgrade the performance of the classifier. We comprehensively evaluate MHCPDP on five open source datasets; our experimental results show that MHCPDP not only has significant improvement in two performance metrics but also overcomes the shortcomings of the conventional HCPDP methods.

Multi-source fast transfer learning algorithm based on support vector machine.

Knowledge in the source domain can be used in transfer learning to help train and classification tasks within the target domain with fewer available data sets. Therefore, given the situation where the target domain contains only a small number of available unlabeled data sets and multi-source domains contain a large number of labeled data sets, a new Multi-source Fast Transfer Learning algorithm based on support vector machine(MultiFTLSVM) is proposed in this paper. Given the idea of multi-source transfer learning, more source domain knowledge is taken to train the target domain learning task to improve classification effect. At the same time, the representative data set of the source domain is taken to speed up the algorithm training process to improve the efficiency of the algorithm. Experimental results on several real data sets show the effectiveness of MultiFTLSVM, and it also has certain advantages compared with the benchmark algorithm.

Predictive Modeling for Machining Power Based on Multi-source Transfer Learning in Metal Cutting

Energy efficiency has become crucial in the metal cutting industry. Machining power has therefore become an important metric because it directly affects the energy consumed during the operation of a machine tool. Attempts to predict machining power using machine learning have relied on the training datasets processed from actual machining data to derive the numerical relationship between process parameters and machining power. However, real fields hardly provide training datasets because of the difficulties in data collection; consequently, traditional learning approaches are ineffective in such data-scarce or -absent environment. This paper proposes a transfer learning approach for the predictive modeling of machining power. The proposed approach creates machining power prediction models by transferring the knowledge acquired from prior machining to the target machining context where machining power data are absent. The proposed approach performs domain adaptation by adding workpiece material properties to the original feature space for accommodating different machining power patterns dependent on the types of workpiece materials. A case study demonstrates that the training datasets obtained from the fabrication of steel and aluminum materials can be successfully used to create the power-predictive models that anticipate machining power for titanium material.

Multi-source cross-project software defect prediction based on deep integration

Cross-project defect prediction (CPDP), training a machine learning model by using training data from other projects, has attracted much attention in recent years. This approach provides a feasible way for small-scale or new developed project with insufficient training data to carry out defect prediction. This paper focus on the bottleneck issue of CPDP, poor accuracy, and propose a deep learning model integration-based approach (MTrADL) for CPDP. This paper consists of two main phases. First, the similarity between target project and source projects is measured by the modified maximum mean discrepancy (MMD) and the top K source projects with high similarity to the target project are selected as training data. Second, for the selected training data, this paper use convolutional neural network (CNN) to build the defect predictor. Each selected training data corresponds to one CNN predictor. Then, multiple predictors are integrated to get the final prediction result. To examine the performance of the proposed approach, this paper conduct experiments on 41 datasets of PROMISE and compare our approach with three state-of-the-art baseline approaches: a training data selection model (TDS), a two-stage transfer learning model (TPTL), and the multi-source transfer learning model (MTrA). The experimental results show that the average F1-score of our approach is 0.76. Across the 41 datasets, on average, MTrADL respectively improves these baseline models by 39.8%, 28.50%, and 10% in terms of F1-score.

Anomaly based novel multi-source unsupervised transfer learning approach for carbon emission centric GDP prediction

Historically, non-renewable fuels are predominant energy sources which drive various industrial and economic activities of a nation. Moreover, it has been well established in literature that carbon emission due to fuel consumption has proportional effect on the economic growth of a nation. There are many war-torn/isolated countries in the world whose gross domestic product (GDP) values are unavailable for certain years in the data repository of the World Bank. Since, hardly any work addresses this research gap of estimating missing GDP of such nations. This paper aims to predict the missing per capita GDP of war-torn/isolated countries using only their carbon emission data. The available data of these countries are not sufficient enough to train a robust predictive model to estimate their missing GDPs. Therefore, this paper proposes multi-source unsupervised transfer learning (UTL), which is an emerging yet under-explored area of transfer learning, to enlarge the training datasets in order to build a robust GDP prediction model. Thus, it exploits multi-source UTL to generate a set of distinct training domains from the collection of carbon emission dataset of various developing countries as well as a mixture of developed and developing countries by suitably detecting and removing the anomalies using isolation forest. To demonstrate the working of the proposed approach, extensive simulations are carried out by considering the carbon emission data sets of six war-torn/isolated countries viz., Afghanistan, Myanmar, Syria, Vietnam, Yemen and Iraq. Here, five different training domains were generated which are then empirically evaluated using three different kind of predictive regression models. The domain with best resulting accuracy is chosen to predict the missing per capita GDP of these war-torn/isolated nations. It is found that the proposed approach performed exceptionally well in increasing the size of relevant training dataset and hence provides a robust GDP prediction model. The strength and weakness of the proposed approach are also discussed by emphasizing the constraints that needs to be adhered while using the proposed UTL approach.

Improving prediction for medical institution with limited patient data: Leveraging hospital-specific data based on multicenter collaborative research network

Background and objectiveClinical decision support assisted by prediction models usually faces the challenges of limited clinical data and a lack of labels when the model is developed with data from a single medical institution. Accordingly, research on multicenter clinical collaborative networks, which can provide external medical data, has received increasing attention. With the increasing availability of machine learning techniques such as transfer learning, leveraging large-scale patient data from multiple hospitals to build data-driven predictive models with clinical application potential provides an alternative solution to address the problem of limited patient data. MethodsA multicenter hybrid semi-supervised transfer learning model (MHSTL) is proposed in this study on the basis of unified common data model to ensure multicenter data standardized representation. Then the hospital-specific features, along with the co-occurrence features across domains, are aligned through a representation learning architecture that is built based on deep neural networks and the newly proposed neural decision forest model. In this process, limited patient data from the target hospital, both labeled and unlabeled, are incorporated during the feature adaptation process, thereby contributing to better model performance. Without patient-level data sharing, the proposed model learning strategy which overcomes feature misalignment and distribution divergence, enables the multi-source transfer learning process in the case of insufficient and unlabeled patient data at target hospital. ResultsThe effectiveness of the proposed transfer learning model was evaluated on a collaborative research network of colorectal cancer patients in the US and China. The results demonstrate that the proposed model can achieve much better performance for predicting target risk with limited resources on patient data than baseline models . Better discrimination and calibration ability are also observed when sufficient labeled data are not available in the target hospital for prognosis prediction tasks . Further exploratory experiments show that the proposed approach exhibits good model generalizability regardless of the data heterogeneity. With the help of the SHapley Additive exPlanations for model interpretation, the effectiveness of incorporating hospital-specific features in the transfer learning model is shown. ConclusionsIn this study, the proposed method can develop prediction models from multiple source hospitals and exhibit good performance by leveraging cross-domain hospital-specific feature information, therefore enhancing the model prediction when applied to single medical institution with limited patient data.

A transfer learning model with multi-source domains for biomedical event trigger extraction

BackgroundAutomatic extraction of biomedical events from literature, that allows for faster update of the latest discoveries automatically, is a heated research topic now. Trigger word recognition is a critical step in the process of event extraction. Its performance directly influences the results of the event extraction. In general, machine learning-based trigger recognition approaches such as neural networks must to be trained on a dataset with plentiful annotations to achieve high performances. However, the problem of the datasets in wide coverage event domains is that their annotations are insufficient and imbalance. One of the methods widely used to deal with this problem is transfer learning. In this work, we aim to extend the transfer learning to utilize multiple source domains. Multiple source domain datasets can be jointly trained to help achieve a higher recognition performance on a target domain with wide coverage events.ResultsBased on the study of previous work, we propose an improved multi-source domain neural network transfer learning architecture and a training approach for biomedical trigger detection task, which can share knowledge between the multi-source and target domains more comprehensively. We extend the ability of traditional adversarial networks to extract common features between source and target domains, when there is more than one dataset in the source domains. Multiple feature extraction channels to simultaneously capture global and local common features are designed. Moreover, under the constraint of an extra classifier, the multiple local common feature sub-channels can extract and transfer more diverse common features from the related multi-source domains effectively. In the experiments, MLEE corpus is used to train and test the proposed model to recognize the wide coverage triggers as a target dataset. Other four corpora with the varying degrees of relevance with MLEE from different domains are used as source datasets, respectively. Our proposed approach achieves recognition improvement compared with traditional adversarial networks. Moreover, its performance is competitive compared with the results of other leading systems on the same MLEE corpus.ConclusionsThe proposed Multi-Source Transfer Learning-based Trigger Recognizer (MSTLTR) can further improve the performance compared with the traditional method, when the source domains are more than one. The most essential improvement is that our approach represents common features in two aspects: the global common features and the local common features. Hence, these more sharable features improve the performance and generalization of the model on the target domain effectively.

Adversarial multi-source transfer learning in healthcare: Application to glucose prediction for diabetic people

Background and objectives: Deep learning has yet to revolutionize general practices in healthcare, despite promising results for some specific tasks. This is partly due to data being in insufficient quantities hurting the training of the models. To address this issue, data from multiple health actors or patients could be combined by capitalizing on their heterogeneity through the use of transfer learning.Methods: To improve the quality of the transfer between multiple sources of data, we propose a multi-source adversarial transfer learning framework that enables the learning of a feature representation that is similar across the sources, and thus more general and more easily transferable. We apply this idea to glucose forecasting for diabetic people using a fully convolutional neural network. The evaluation is done by exploring various transfer scenarios with three datasets characterized by their high inter and intra variability.Results: While transferring knowledge is beneficial in general, we show that the statistical and clinical accuracies can be further improved by using of the adversarial training methodology, surpassing the current state-of-the-art results. In particular, it shines when using data from different datasets, or when there is too little data in an intra-dataset situation. To understand the behavior of the models, we analyze the learnt feature representations and propose a new metric in this regard. Contrary to a standard transfer, the adversarial transfer does not discriminate the patients and datasets, helping the learning of a more general feature representation.Conclusion: The adversarial training framework improves the learning of a general feature representation in a multi-source environment, enhancing the knowledge transfer to an unseen target. The proposed method can help improve the efficiency of data shared by different health actors in the training of deep models.

A novel active multi-source transfer learning algorithm for time series forecasting

In Time Series Forecasting (TSF), researchers usually assume that there is enough training data can be obtained, with the old a`nd new data satisfying the same distribution. However, time series data always produces some time-varying characteristics over time, which will lead to relatively large differences between old and new data. As we all know, single-source TSF Transfer Learning (TL) faces the problem of negative transfer. Addressing this issue, this paper proposes a new Multi-Source TL algorithm, abbreviated as the MultiSrcTL algorithm, and a novel Active Multi-Source Transfer Learning, abbreviated as the AcMultiSrcTL algorithm, with the latter one integrating Multi-Source TL with Active Learning (AL), and taking the former one as its sub-algorithm. We introduce domain adaptation theory into this work, and analyze the expected target risk of TSF under the multi-source setting, accordingly. For the development of MultiSrcTL, we make full use of source similarity and domain dependability, using the Maximum Mean Discrepancy statistical indicator to measure the similarity between domains, so as to promote better transfer. A domain relation matrix is constructed to describe the relationship between source domains, so that the source-source and source-target relations are adequately considered. In the design of AcMultiSrcTL, Kullback-Leibler divergence is used to measure the similarity of related indicators to select the appropriate source domain. The uncertainty sampling method and the distribution match weighting technique are integrated, obtaining a new sample selection scheme. The empirical results on six benchmark datasets demonstrate the applicability and effectiveness of the two proposed algorithms for multi-source TSF TL.

Inter-subject pattern analysis for multivariate group analysis of functional neuroimaging. A unifying formalization

Background and objectiveIn medical imaging, population studies have to overcome the differences that exist between individuals to identify invariant image features that can be used for diagnosis purposes. In functional neuroimaging, an appealing solution to identify neural coding principles that hold at the population level is inter-subject pattern analysis, i.e. to learn a predictive model on data from multiple subjects and evaluate its generalization performance on new subjects. Although it has gained popularity in recent years, its widespread adoption is still hampered by the blatant lack of a formal definition in the literature. In this paper, we precisely introduce the first principled formalization of inter-subject pattern analysis targeted at multivariate group analysis of functional neuroimaging. MethodsWe propose to frame inter-subject pattern analysis as a multi-source transductive transfer question, thus grounding it within several well defined machine learning settings and broadening the spectrum of usable algorithms. We describe two sets of inter-subject brain decoding experiments that use several open datasets: a magneto-encephalography study with 16 subjects and a functional magnetic resonance imaging paradigm with 100 subjects. We assess the relevance of our framework by performing model comparisons, where one brain decoding model exploits our formalization while others do not. ResultsThe first set of experiments demonstrates the superiority of a brain decoder that uses subject-by-subject standardization compared to state of the art models that use other standardization schemes, making the case for the interest of the transductive and the multi-source components of our formalization The second set of experiments quantitatively shows that, even after such transformation, it is more difficult for a brain decoder to generalize to new participants rather than to new data from participants available in the training phase, thus highlighting the transfer gap that needs to be overcome. ConclusionThis paper describes the first formalization of inter-subject pattern analysis as a multi-source transductive transfer learning problem. We demonstrate the added value of this formalization using proof-of-concept experiments on several complementary functional neuroimaging datasets. This work should contribute to popularize inter-subject pattern analysis for functional neuroimaging population studies and pave the road for future methodological innovations.

Practical Multisource Transfer Regression With Source-Target Similarity Captures.

A key challenge in many applications of multisource transfer learning is to explicitly capture the diverse source-target similarities. In this article, we are concerned with stretching the set of practical approaches based on Gaussian process (GP) models to solve multisource transfer regression problems. Precisely, we first investigate the feasibility and performance of a family of transfer covariance functions that represent the pairwise similarity of each source and the target domain. We theoretically show that using such a transfer covariance function for general GP modeling can only capture the same similarity coefficient for all the sources, and thus may result in unsatisfactory transfer performance. This outcome, together with the scalability issues of a single GP based approach, leads us to propose TCMSStack , an integrated framework incorporating a separate transfer covariance function for each source and stacking. Contrary to typical stacking approaches, TCMSStack learns the source-target similarity in each base GP model by considering the dependencies of the other sources along the process. We introduce two instances of the proposed TCMSStack . Extensive experiments on one synthetic and two real-world data sets, with learning settings up to 11 sources for the latter, demonstrate the effectiveness of our approach.

Calibrating EEG features in motor imagery classification tasks with a small amount of current data using multisource fusion transfer learning

Background and objectiveThe identification of motor intention by analyzing electroencephalogram (EEG) signals has become an important issue in brain-computer interface (BCI) applications. Since the individual differences of EEG are large, it’s hard to design a universal BCI system, which means the labeled data from current user have to be collected for calibration. We propose a multi-source fusion transfer learning algorithm for this calibration, which only needs a small amount of new data from the current user. MethodsThe core idea of proposed algorithm is a two-step calibration based on transfer learning, which works on the levels of both subjects and features. The spatial covariance matrices represented in the Riemannian space are used to characterize the multi-channel EEG signals. Source subjects whose features are similar to the current user are selected as the transfer source via a Riemannian geometry alignment algorithm. Their features obtained from the Riemannian tangent space, as well as the ones of the current user, are fused and calibrated by a balanced distribution adaptation algorithm. Finally, a classification model could be trained for the current user with only a few new training data. ResultsValidated on both the online ERD-BCI database and the BCI Competition IV 2a database for motor imagery tasks, experimental results show that our method outperforms the state-of-the-art methods. When the new training data from current user is insufficient, the classification accuracy and kappa values compared with the outstanding counterparts have advantages of 3.30 % and 0.05 respectively on the online ERD-BCI database, 3.95 % and 0.03 respectively on the BCI Competition IV 2a database. Even when the new training data from current user is sufficient, the classification accuracy and kappa values of our proposed method also outperforms the counterparts with highest performance by 2.37 % and 0.05 respectively on the online ERD-BCI database, 2.25 % and 0.03 respectively on the BCI Competition IV 2a database. ConclusionsThe algorithm provides a new idea of solution for calibrating EEG features. It is expected to contribute to the practical applications of BCI.

A Transfer Learning Framework for RSVP-based Brain Computer Interface.

Rapid Serial Visual Presentation (RSVP)-based Brain-Computer Interface (BCI) is an efficient information detection technology by detecting event-related brain responses evoked by target visual stimuli. However, a time-consuming calibration procedure is needed before a new user can use this system. Thus, it is important to reduce calibration efforts for BCI applications. In this paper, we collect an RSVP-based electroencephalogram (EEG) dataset, which includes 11 subjects. The experimental task is image retrieval. Also, we propose a multi-source transfer learning framework by utilizing data from other subjects to reduce the data requirement on the new subject for training the model. A source-selection strategy is firstly adopted to avoid negative transfer. And then, we propose a transfer learning network based on domain adversarial training. The convolutional neural network (CNN)-based network is designed to extract common features of EEG data from different subjects, while the discriminator tries to distinguish features from different subjects. In addition, a classifier is added for learning semantic information. Also, conditional information and gradient penalty are added to enable stable training of the adversarial network and improve performance. The experimental results demonstrate that our proposed method outperforms a series of state-of-the-art and baseline approaches.