Transfer Learning

Transfer learning aims at improving the performance of target learners on target domains by transferring the knowledge contained in different but related source domains. In this way, the dependence on a large number of target-domain data can be reduced for constructing target learners. Due to the wide application prospects, transfer learning has become a popular and promising area in machine learning. Although there are already some valuable and impressive surveys on transfer learning, these surveys introduce approaches in a relatively isolated way and lack the recent advances in transfer learning. Due to the rapid expansion of the transfer learning area, it is both necessary and challenging to comprehensively review the relevant studies. This survey attempts to connect and systematize the existing transfer learning research studies, as well as to summarize and interpret the mechanisms and the strategies of transfer learning in a comprehensive way, which may help readers have a better understanding of the current research status and ideas. Unlike previous surveys, this survey article reviews more than 40 representative transfer learning approaches, especially homogeneous transfer learning approaches, from the perspectives of data and model. The applications of transfer learning are also briefly introduced. In order to show the performance of different transfer learning models, over 20 representative transfer learning models are used for experiments. The models are performed on three different data sets, that is, Amazon Reviews, Reuters-21578, and Office-31, and the experimental results demonstrate the importance of selecting appropriate transfer learning models for different applications in practice.

Cross-lingual sentiment classification (CLSC) aims to leverage rich-labeled resources in the source language to improve prediction models of a resource-scarce domain in the target language. Existing feature representation learning based approaches try to minimize the difference of latent features between different domains by exact alignment, which is achieved by either one-to-one topic alignment or matrix projection. Exact alignment, however, restricts the representation flexibility, and further degrades model performances on cross-lingual sentiment classification tasks if the distribution difference between two language domains is large. On the other hand, most previous studies proposed document-level models or ignored sentiment polarities of topics that might lead to insufficient learning of latent features. To solve the above problems, we propose a coarse alignment mechanism to enhance model’s representation by a group-to-group topic alignment into an aspect-level fine-grained model. First, we propose an unsupervised aspect, opinion and sentiment unification model (AOS), which tri-models aspects, opinions and sentiments of reviews from different domains and helps capture more accurate latent feature representation by a coarse alignment mechanism. To further boost AOS, we propose ps-AOS, a partial supervised AOS model, in which labeled source language data help minimize the difference of feature representations between two language domains with the help of logistics regression. Last, an Expectation-Maximization framework with Gibbs sampling is then proposed to optimize our model. Extensive experiments on various multilingual product review data sets show that ps-AOS significantly outperforms various kinds of state-of-the-art baselines.

For a target task where labeled data is unavailable, domain adaptation can transfer a learner from a different source domain. Previous deep domain adaptation methods mainly learn a global domain shift, i.e., align the global source and target distributions without considering the relationships between two subdomains within the same category of different domains, leading to unsatisfying transfer learning performance without capturing the fine-grained information. Recently, more and more researchers pay attention to Subdomain Adaptation which focuses on accurately aligning the distributions of the relevant subdomains. However, most of them are adversarial methods which contain several loss functions and converge slowly. Based on this, we present Deep Subdomain Adaptation Network (DSAN) which learns a transfer network by aligning the relevant subdomain distributions of domain-specific layer activations across different domains based on a local maximum mean discrepancy (LMMD). Our DSAN is very simple but effective which does not need adversarial training and converges fast. The adaptation can be achieved easily with most feed-forward network models by extending them with LMMD loss, which can be trained efficiently via back-propagation. Experiments demonstrate that DSAN can achieve remarkable results on both object recognition tasks and digit classification tasks. Our code will be available at: https://github.com/easezyc/deep-transfer-learning.

Visual aesthetics of products plays an important role in the decision process when purchasing appearance-first products, e.g., clothes. Indeed, user’s aesthetic preference, which serves as a personality trait and a basic requirement, is domain independent and could be used as a bridge between domains for knowledge transfer. However, existing work has rarely considered the aesthetic information in product images for cross-domain recommendation. To this end, in this paper, we propose a new deep Aesthetic Cross-Domain Networks (ACDN), in which parameters characterizing personal aesthetic preferences are shared across networks to transfer knowledge between domains. Specifically, we first leverage an aesthetic network to extract aesthetic features. Then, we integrate these features into a cross-domain network to transfer users’ domain independent aesthetic preferences. Moreover, network cross-connections are introduced to enable dual knowledge transfer across domains. Finally, the experimental results on real-world datasets show that our proposed model ACDN outperforms benchmark methods in terms of recommendation accuracy.

With the explosive growth of the e-commerce industry, detectingonline transaction fraud in real-world applications has become increasingly important to the development of e-commerce platforms. The sequential behavior history of users provides useful information in differentiating fraudulent payments from regular ones. Recently, some approaches have been proposed to solve this sequence-based fraud detection problem. However, these methods usually suffer from two problems: the prediction results are difficult to explain and the exploitation of the internal information of behaviors is insufficient. To tackle the above two problems, we propose a Hierarchical Explainable Network (HEN) to model users’ behavior sequences, which could not only improve the performance of fraud detection but also make the inference process interpretable.

Meanwhile, as e-commerce business expands to new domains, e.g., new countries or new markets, one major problem for modeling user behavior in fraud detection systems is the limitation of data collection, e.g., very few data/labels available. Thus, in this paper, we further propose a transfer framework to tackle the cross-domain fraud detection problem, which aims to transfer knowledge from existing domains (source domains) with enough and mature data to improve the performance in the new domain (target domain). Our proposed method is a general transfer framework that could not only be applied upon HEN but also various existing models in the Embedding & MLP paradigm.

By utilizing data from a world-leading cross-border e-commerce platform, we conduct extensive experiments in detecting card-stolen transaction frauds in different countries to demonstrate the superior performance of HEN. Besides, based on 90 transfer task experiments, we also demonstrate that our transfer framework could not only con?tribute to the cross-domain fraud detection task with HEN, but also be universal and expandable for various existing models. Moreover, HEN and the transfer framework form three-level attention which greatly increases the explainability of the detection results.

Domain adaptation tasks such as cross-domain sentiment classification aim to utilize existing labeled data in the source domain and unlabeled or few labeled data in the target domain to improve the performance in the target domain via reducing the shift between the data distributions. Existing cross-domain sentiment classification methods need to distinguish pivots, i.e., the domain-shared sentiment words, and non-pivots, i.e., the domain-specific sentiment words, for excellent adaptation performance.

In this paper, we first design a Category Attention Network (CAN), and then propose a model named CAN-CNN to integrate CAN and a Convolutional Neural Network (CNN). On the one hand, the model regards pivots and non-pivots as unified category attribute words and can automatically capture them to improve the domain adaptation performance; on the other hand, the model makes an attempt at interpretability to learn the transferred category attribute words. Specifically, the optimization objective of our model has three different components: 1) the supervised classification loss; 2) the distributions loss of category feature weights; 3) the domain invariance loss.

Finally, the proposed model is evaluated on three public sentiment analysis datasets and the results demonstrate that CAN-CNN can outperform other various baseline methods.

While Unsupervised Domain Adaptation (UDA) algorithms, i.e., there are only labeled data from source domains, have been actively studied in recent years, most algorithms and theoretical results focus on Single-source Unsupervised Domain Adaptation (SUDA). However, in the practical scenario, labeled data can be typically collected from multiple diverse sources, and they might be different not only from the target domain but also from each other. Thus, domain adapters from multiple sources should not be modeled in the same way. Recent deep learning based Multi-source Unsupervised Domain Adaptation (MUDA) algorithms focus on extracting common domain-invariant representations for all domains by aligning distribution of all pairs of source and target domains in a common feature space. However, it is often very hard to extract the same domain-invariant representations for all domains in MUDA. In addition, these methods match distributions without considering domain-specific decision boundaries between classes. To solve these problems, we propose a new framework with two alignment stages for MUDA which not only respectively aligns the distributions of each pair of source and target domains in multiple specific feature spaces, but also aligns the outputs of classifiers by utilizing the domain-specific decision boundaries. Extensive experiments demonstrate that our method can achieve remarkable results on popular benchmark datasets for image classification.

The main challenge of cross-domain text classification is to train a classifier in a source domain while applying it to a different target domain. Many transfer learning-based algorithms, for example, dual transfer learning, triplex transfer learning, etc., have been proposed for cross-domain classification, by detecting a shared low-dimensional feature representation for both source and target domains. These methods, however, often assume that the word clusters matrix or the clusters association matrix as knowledge transferring bridges are exactly the same across different domains, which is actually unrealistic in real-world applications and, therefore, could degrade classification performance. In light of this, in this paper, we propose a softly associative transfer learning algorithm for cross-domain text classification. Specifically, we integrate two non-negative matrix tri-factorizations into a joint optimization framework, with approximate constraints on both word clusters matrices and clusters association matrices so as to allow proper diversity in knowledge transfer, and with another approximate constraint on class labels in source domains in order to handle noisy labels. An iterative algorithm is then proposed to solve the above problem, with its convergence verified theoretically and empirically. Extensive experimental results on various text datasets demonstrate the effectiveness of our algorithm, even with the presence of abundant state-of-the-art competitors.

Transfer learning has gained a lot of attention and interest in the past decade. One crucial research issue in transfer learning is how to find a good representation for instances of different domains such that the divergence between domains can be reduced with the new representation. Recently, deep learning has been proposed to learn more robust or higher-level features for transfer learning. In this article, we adapt the autoencoder technique to transfer learning and propose a supervised representation learning method based on double encoding-layer autoencoder. The proposed framework consists of two encoding layers: one for embedding and the other one for label encoding. In the embedding layer, the distribution distance of the embedded instances between the source and target domains is minimized in terms of KL-Divergence. In the label encoding layer, label information of the source domain is encoded using a softmax regression model. Moreover, to empirically explore why the proposed framework can work well for transfer learning, we propose a new effective measure based on autoencoder to compute the distribution distance between different domains. Experimental results show that the proposed new measure can better reflect the degree of transfer difficulty and has stronger correlation with the performance from supervised learning algorithms (e.g., Logistic Regression), compared with previous ones, such as KL-Divergence and Maximum Mean Discrepancy. Therefore, in our model, we have incorporated two distribution distance measures to minimize the difference between source and target domains in the embedding representations. Extensive experiments conducted on three real-world image datasets and one text data demonstrate the effectiveness of our proposed method compared with several state-of-the-art baseline methods.

Cross-domain text classification aims at building a classifier for a target domain which leverages data from both source and target domain. One promising idea is to minimize the feature distribution differences of the two domains. Most existing studies explicitly minimize such differences by an exact alignment mechanism (aligning features by one-to-one feature alignment, projection matrix etc.). Such exact alignment, however, will restrict models’ learning ability and will further impair models’ performance on classification tasks when the semantic distributions of different domains are very different. To address this problem, we propose a novel group alignment which aligns the semantics at group level. In addition, to help the model learn better semantic groups and semantics within these groups, we also propose a partial supervision for model’s learning in source domain. To this end, we embed the group alignment and a partial supervision into a cross-domain topic model, and propose a Cross-Domain Labeled LDA (CDL-LDA). On the standard 20Newsgroup and Reuters dataset, extensive quantitative (classification, perplexity etc.) and qualitative (topic detection) experiments are conducted to show the effectiveness of the proposed group alignment and partial supervision.

Collaborative filtering is an effective and widely used recommendation approach by applying the user-item rating matrix for recommendations, however, which usually suffers from cold-start and sparsity problems. To address these problems, hybrid methods are proposed to incorporate auxiliary information such as user/item profiles to collaborative filtering models; Cross-domain recommendation systems add a new dimension to solve these problems by leveraging ratings from other domains to improve recommendation performance. Among these methods, deep neural network based recommendation systems achieve excellent performance due to their excellent ability in learning powerful representations. However, these cross-domain recommendation systems based on deep neural network rarely consider the uncertainty of weights. Therefore, they maybe lack of calibrated probabilistic predictions and make overly confident decisions. Along this line, we propose a general cross-domain recommendation framework via Bayesian neural network to incorporate auxiliary information, which takes advantage of both the hybrid recommendation methods and the cross-domain recommendation systems. Specifically, our framework consists of two kinds of neural networks, one to learn the low dimensional representation from the one-hot codings of users/items, while the other one is to project the auxiliary information of users/items into another latent space. The final rating is produced by integrating the latent representations of the one-hot codings of users/items and the auxiliary information of users/items. The latent representations of users learnt from ratings and auxiliary information are shared across different domains for knowledge transfer. Moreover, we capture the uncertainty in all weights by representing weights with Gaussian distributions to make calibrated probabilistic predictions. We have done extensive experiments on real-world data sets to verify the effectiveness of our framework.

Recent studies in psychology suggest that novelty-seeking trait is highly related to consumer behavior, which has a profound business impact on online recommendation. This paper studies the problem of mining novelty seeking trait across domains to improve the recommendation performance in target domain. We propose an efficient model, CDNST, which significantly improves the recommendation performance by transferring the knowledge from auxiliary source domain. We conduct extensive experiments on three domain datasets crawled from Douban (www.douban.com) to demonstrate the effectiveness of the proposed model. Moreover, we find that the property of sequential data affects the performance of CDNST.

In the past decade, there have been a large number of transfer learning algorithms proposed for various real-world applications. However, most of them are vulnerable to negative transfer1 since their performance is even worse than traditional supervised models. Aiming at more robust transfer learning models, we propose an ENsemble framework of anCHOR adapters (ENCHOR for short), in which an anchor adapter adapts the features of instances based on their similarities to a specif c anchor (i.e., a selected instance). Specif cally, the more similar to the anchor instance, the higher degree of the original feature of an instance remains unchanged in the adapted representation, and vice versa. This adapted representation for the data actually expresses the local structure around the corresponding anchor, and then any transfer learning method can be applied to this adapted representation for a prediction model, which focuses more on the neighborhood of the anchor. Next, based on multiple anchors, multiple anchor adapters can be built and combined into an ensemble for f nal output. Additionally, we develop an effective measure to select the anchors for ensemble building to achieve further performance improvement. Extensive experiments on hundreds of text classif cation tasks are conducted to demonstrate the effectiveness of ENCHOR. The results show that: when traditional supervised models perform poorly, ENCHOR (based on only 8 selected anchors) achieves 6% − 13% increase in terms of average accuracy compared with the state-of-the-art methods, and it greatly alleviates negative transfer.

Transfer learning has attracted a lot of attention in the past decade. One crucial research issue in transfer learning is how to find a good representation for instances of different domains such that the divergence between domains can be reduced with the new representation. Recently, deep learning has been proposed to learn more robust or higherlevel features for transfer learning. However, to the best of our knowledge, most of the previous approaches neither minimize the difference between domains explicitly nor encode label information in learning the representation. In this paper, we propose a supervised representation learning method based on deep autoencoders for transfer learning. The proposed deep autoencoder consists of two encoding layers: an embedding layer and a label encoding layer. In the embedding layer, the distance in distributions of the embedded instances between the source and target domains is minimized in terms of KL-Divergence. In the label encoding layer, label information of the source domain is encoded using a softmax regression model. Extensive experiments conducted on three real-world image datasets demonstrate the effectiveness of our proposed method compared with several state-of-theart baseline methods.

In this study, we consider an ensemble problem in which we combine outputs coming from models developed in the supervised and unsupervised modes. By jointly considering the grouping results coming from unsupervised models we aim to improve the classification accuracy of supervised model ensemble. Here, we formulate the ensemble task as an Unconstrained Probabilistic Embedding (UPE) problem. Specifically, we assume both objects and classes/clusters have latent coordinates without constraints in a D-dimensional Euclidean space, and consider the mapping from the embedded space into the space of model results as a probabilistic generative process. A solution to this embedding can be obtained using the quasi-Newton method, which makes objects and classes/clusters with high cooccurrence weights are embedded close. Then, prediction is determined by taking the distances between the object and the classes in the embedded space. We demonstrate the benefits of this unconstrained embedding method by running extensive and systematic experiments on real-world datasets. Furthermore, we conduct experiments to investigate how the quality and the number of clustering models affect the performance of this ensemble method. We also show the robustness of the proposed model.

Knowledge transfer from multiple source domains to a target domain is crucial in transfer learning. Most existing methods are focused on learning weights for different domains based on the similarities between each source domain and the target domain or learning more precise classifiers from the source domain data jointly by maximizing their consensus of predictions on the target domain data. However, these methods only consider measuring similarities or building classifiers on the original data space, and fail to discover a more powerful feature representation of the data when transferring knowledge from multiple source domains to the target domain. In this paper, we propose a new framework for transfer learning with multiple source domains. Specifically, in the proposed framework, we adopt autoencoders to construct a feature mapping from an original instance to a hidden representation, and train multiple classifiers from the source domain data jointly by performing an entropy-based consensus regularizer on the predictions on the target domain. Based on the framework, a particular solution is proposed to learn the hidden representation and classifiers simultaneously. Experimental results on image and text real-world datasets demonstrate the effectiveness of our proposed method compared with state-of-the-art methods.

Cross-domain learning targets at leveraging the knowledge from source domains to train accurate models for the test data from target domains with different but related data distributions. To tackle the challenge of data distribution difference in terms of raw features, previous works proposed to mine high-level concepts (e.g., word clusters) across data domains, which shows to be more appropriate for classification. However, all these works assume that the same set of concepts are shared in the source and target domains in spite that some distinct concepts may exist only in one of the data domains. Thus, we need a general framework, which can incorporate both shared and distinct concepts, for cross-domain classification. To this end, we develop a probabilistic model, by which both the shared and distinct concepts can be learned by the EM process which optimizes the data likelihood. To validate the effectiveness of this model we intentionally construct the classification tasks where the distinct concepts exist in the data domains. The systematic experiments demonstrate the superiority of our model over all compared baselines, especially on those much more challenging tasks.

The distribution difference among multiple domains has been exploited for cross-domain text categorization in recent years. Along this line, we show two new observations in this study. First, the data distribution difference is often due to the fact that different domains use different index words to express the same concept. Second, the association between the conceptual feature and the document class can be stable across domains. These two observations actually indicate the distinction and commonality across domains. Inspired by the above observations, we propose a generative statistical model, named Collaborative Dual-PLSA (CD-PLSA), to simultaneously capture both the domain distinction and commonality among multiple domains. Different from Probabilistic Latent Semantic Analysis (PLSA) with only one latent variable, the proposed model has two latent factors y and z, corresponding to word concept and document class, respectively. The shared commonality intertwines with the distinctions over multiple domains, and is also used as the bridge for knowledge transformation. An Expectation Maximization (EM) algorithm is developed to solve the CD-PLSA model, and further its distributed version is exploited to avoid uploading all the raw data to a centralized location and help to mitigate privacy concerns. After the training phase with all the data from multiple domains we propose to refine the immediate outputs using only the corresponding local data. In summary, we propose a two-phase method for cross-domain text classification, the first phase for collaborative training with all the data, and the second step for local refinement. Finally, we conduct extensive experiments over hundreds of classification tasks with multiple source domains and multiple target domains to validate the superiority of the proposed method over existing state-of-the-art methods of supervised and transfer learning. It is noted to mention that as shown by the experimental results CD-PLSA for the collaborative training is more tolerant of distribution differences, and the local refinement also gains significant improvement in terms of classification accuracy.

Ensemble learning with output from multiple supervised and unsupervised models aims to improve the classification accuracy of supervised model ensemble by jointly considering the grouping results from unsupervised models. In this paper we cast this ensemble task as an unconstrained probabilistic embedding problem. Specifically, we assume both objects and classes/clusters have latent coordinates without constraints in a D-dimensional Euclidean space, and consider the mapping from the embedded space into the space of results from supervised and unsupervised models as a probabilistic generative process. The prediction of an object is then determined by the distances between the object and the classes in the embedded space. A solution of this embedding can be obtained using the quasi-Newton method, resulting in the objects and classes/clusters with high co-occurrence weights being embedded close. We demonstrate the benefits of this unconstrained embedding method by three real applications.

Classification across different domains studies how to adapt a learning model from one domain to another domain which shares similar data characteristics. While there are a number of existing works along this line, many of them are only focused on learning from a single source domain to a target domain. In particular, a remaining challenge is how to apply the knowledge learned from multiple source domains to a target domain. Indeed, data from multiple source domains can be semantically related, but have different data distributions. It is not clear how to exploit the distribution differences among multiple source domains to boost the learning performance in a target domain. To that end, in this paper, we propose a consensus regularization framework for learning from multiple source domains to a target domain. In this framework, a local classifier is trained by considering both local data available in one source domain and the prediction consensus with the classifiers learned from other source domains. Moreover, we provide a theoretical analysis as well as an empirical study of the proposed consensus regularization framework. The experimental results on text categorization and image classification problems show the effectiveness of this consensus regularization learning method. Finally, to deal with the situation that the multiple source domains are geographically distributed, we also develop the distributed version of the proposed algorithm, which avoids the need to upload all the data to a centralized location and helps to mitigate privacy concerns.

The distribution difference among multiple data domains has been considered for the cross-domain text classification problem. In this study, we show two new observations along this line. First, the data distribution difference may come from the fact that different domains use different key words to express the same concept. Second, the association between this conceptual feature and the document class may be stable across domains. These two issues are actually the distinction and commonality across data domains.
Inspired by the above observations, we propose a generative statistical model, named Collaborative Dual-PLSA (CD-PLSA), to simultaneously capture both the domain distinction and commonality among multiple domains. Different from Probabilistic Latent Semantic Analysis (PLSA) with only one latent variable, the proposed model has two latent factors y and z, corresponding to word concept and document class respectively. The shared commonality intertwines with the distinctions over multiple domains, and is also used as the bridge for knowledge transformation. We exploit an Expectation Maximization (EM) algorithm to learn this model, and also propose its distributed version to handle the situation where the data domains are geographically separated from each other. Finally, we conduct extensive experiments over hundreds of classification tasks with multiple source domains and multiple target domains to validate the superiority of the proposed CD-PLSA model over existing state-of-the-art methods of supervised and transfer learning. In particular, we show that CD-PLSA is more tolerant of distribution differences.

Cross-domain text categorization targets on adapting the knowledge learnt from a labeled source-domain to an unlabeled target-domain, where the documents from the source and target domains are drawn from different distributions. However, in spite of the different distributions in raw word features, the associations between word clusters (conceptual features) and document classes may remain stable across different domains. In this paper, we exploit these unchanged associations as the bridge of knowledge transformation from the source domain to the target domain by the nonnegative matrix tri-factorization. Specifically, we formulate a joint optimization framework of the two matrix tri-factorizations for the source and target domain data respectively, in which the associations between word clusters and document classes are shared between them. Then, we give an iterative algorithm for this optimization and theoretically show its convergence. The comprehensive experiments show the effectiveness of this method. In particular, we show that the proposed method can deal with some difficult scenarios where baseline methods usually do not perform well.