Vehicle re-identification(Re-ID) is a critical task in Intelligent Transportation Systems, which aims to accurately identify vehicles across different viewpoints and time through visual analysis. However, existing Re-ID methods often struggle in truck scenarios due to the heterogeneity between the truck front and carriage features, dynamic carriage changes, and missing carriage information. To address these issues, we propose a two-stage truck re-identification framework based on domain adversarial learning and distillation learning (T2RIAD). The framework adopts a two-stage training strategy. In the first stage, for known carriage types, domain adversarial learning is employed to suppress the interference of dynamic changes in carriage states. Meanwhile, multi-teacher distillation is used to align and unify feature distributions across different carriage types by enforcing local similarity distillation and global logit consistency distillation, thereby guiding the student model to learn more stable and discriminative feature representations. In the second stage, for unknown carriage type, a dynamic cross-type sampling mechanism is introduced to enhance the generalization ability of the student model. Extensive experiments on two public datasets demonstrate the superiority of our T2RIAD over other state-of-the-art methods.

Deep learning-based object detection models have achieved remarkable performance, but their training often relies on extensive data augmentation due to the high cost of manual annotations. Despite significant advancements, existing data augmentation methods face challenges in balancing data diversity and realism. Traditional augmentation techniques, such as copy-paste and image mixing, frequently produce images with limited realism, which may hinder the model’s adaptability to real-world scenarios. On the other hand, generative augmentation methods often depend on visual priors from the original data to ensure the realism of generated images. This reliance constrains the diversity of the generated data and limits their ability to comprehensively represent the feature distribution of the target domain. To address this issue, we propose a novel data augmentation framework for object detection that leverages instance transformation between categories using a diffusion model. This method quantifies semantic similarity through vision-language models and explores feature gap between categories, combining latent space diffusion models to achieve multimodal controllable generation. While preserving the contextual features of the original scene, the target object is replaced with an instance of a semantically similar category, and the features of the original object and the replacement target are fused in the latent space to generate augmented samples with mixed features. Through cross-modal semantic alignment and posterior filtering mechanisms, the quality of the generated data is ensured, producing training samples that expand data distribution diversity while maintaining both visual plausibility and discriminative ambiguity. Experimental evaluations conducted on the COCO dataset and the Road Damage Dataset under few-shot settings demonstrate the effectiveness and generalizability of the proposed approach in improving object detection performance.

Vehicle detection is an important task in intelligent transportation, which can provides services for road condition monitoring and highway toll collection systems. However existing detection methods are limited by two aspects. First, there is a difference between the model knowledge pre-trained on large-scale datasets and the knowledge required for target task. Second, most detection models follow the pattern of single-source learning, which limits the learning ability. To address these problems, we propose a Multi Self-supervised Pre-finetuned Transformer Fusion (MSPTF) network, consisting of two steps: self-supervised pre-finetuning domain knowledge learning and multi-model fusion target task learning. In the first step, we introduced self-supervised learning methods into transformer pre-finetuning to reduce data costs and alleviate knowledge gap. In the second step, we take feature information differences between different model architectures and pre-finetuning tasks into account and propose Multi-model Semantic Consistency Cross-attention Fusion (MSCCF) network to combine different transformer features by considering channel semantic consistency and feature vector semantic consistency, which obtain more complete and proper fusion features for detection task. We experimented the proposed method on two vehicle detection datasets and achieved 1.1%, 5.5% improvement compared with baseline and 0.7%, 1.8% compared with sota, which proved the effectiveness of our method.Note to Practitioners—Vehicle detection can provide basic services for many intelligent transportation scenarios, but the results of training and adjusting existing models in vehicle detection scenarios with limited data are often not ideal. This article proposes MSPTF to improve model performance from two aspects. First, we randomly collect additional vehicle pictures, and train the model without the need for annotation through annotation-free methods such as image feature extraction and reconstruction, so that the model can be trained on more vehicle-related pictures with low data cost. In addition, we trained multiple models in the above way and proposed a fusion method to combine the knowledge of multiple models for vehicle detection. The results show that our Our model is able to learn from additional unlabeled vehicle images as well as multiple trained models information to achieve better vehicle detection.

In the domain of traffic safety and road maintenance, precise detection of road damage is crucial for ensuring safe driving and prolonging road durability. However, current methods often fall short due to limited data. Prior attempts have used Generative Adversarial Networks to generate damage with diverse shapes and manually integrate it into appropriate positions. However, the problem has not been well explored and is faced with two challenges. First, they only enrich the location and shape of damage while neglect the diversity of severity levels, and the realism still needs further improvement. Second, they require a significant amount of manual effort. To address these challenges, we propose an innovative approach. In addition to using GAN to generate damage with various shapes, we further employ texture synthesis techniques to extract road textures. These two elements are then mixed with different weights, allowing us to control the severity of the synthesized damage, which are then embedded back into the original images via Poisson blending. Our method ensures both richness of damage severity and a better alignment with the background. To save labor costs, we leverage structural similarity for automated sample selection during embedding. Each augmented data of an original image contains versions with varying severity levels. We implement a straightforward screening strategy to mitigate distribution drift. Experiments are conducted on a public road damage dataset. The proposed method not only eliminates the need for manual labor but also achieves remarkable enhancements, improving the mAP by 4.1% and the F1-score by 4.5%.

The goal of computer-aided diagnosis is to predict patient’s diseases based on patient’s clinical data. The development of deep learning technology provides new help for clinical diagnosis. In this paper, we propose a new sequence generation model for multi-disease diagnosis prediction based on numerical feature transformation. Our model simultaneously uses patient’s laboratory test results and clinical text as input to diagnose and predict the disease that the patient may have. According to medical knowledge, our model can transform numerical features into descriptive text features, thereby enriching the semantic information of clinical texts. Besides, our model uses attention-based sequence generation methods to achieve the diagnosis of multiple diseases and better utilizes the correlation information between multiple diseases. We evaluate our model’s performance on a dataset of respiratory diseases from the real world, and experimental results show that our model’s accuracy reaches 42.75%, and the [Formula: see text] score reaches 65.65%, which is better than many other methods. It is suitable for the accurate diagnosis of multiple diseases.

Distributed deep learning has been proposed to train large scale neural network models with huge amounts of datasets using multiple workers. Since workers need to frequently communicate with each other to exchange gradients for updating parameters, communication overhead is always a major challenge in distributed deep learning. To cope with this challenge, gradient compression has been used to reduce the amount of data to be exchanged. However, existing compression methods, including both gradient quantization and gradient sparsification, either hurt model performance significantly or suffer from inefficient compression. In this paper, we propose a novel approach, called Standard Deviation based Adaptive Gradient Compression (SDAGC), which can simultaneously achieve model training with low communication overhead and high model performance in synchronous training. SDAGC uses the standard deviation of gradients in each layer of the neural network to dynamically calculate a suitable threshold according to the training process. Moreover, several associated methods, including residual gradient accumulation, local gradient clipping, adaptive learning rate revision and momentum compensation, are integrated to guarantee the convergence of the model. We verify the performance of SDAGC on various machine learning tasks: image classification, language modeling and speech recognition. The experiment results show that, compared with other existing works, SDAGC can achieve a gradient compression ratio from 433× to 2021× with similar or even better accuracy.

Aspect-based sentiment analysis includes opinion aspect extraction and aspect sentiment classification. Researchers have attempted to discover the relationship between these two sub-tasks and have proposed the joint model for solving aspect-based sentiment analysis. However, they ignore a phenomenon: aspect boundary label and sentiment label of the same word can correct each other. To exploit this phenomenon, we propose a novel deep learning model named the label correction model. Specifically, given an input sentence, our model first predicts the aspect boundary label sequence and sentiment label sequence, then re-predicts the aspect boundary (sentiment) label sequence using the embeddings of the previously predicted sentiment (aspect boundary) label. The goal of the re-prediction operation (can be repeated multiple times) is to use the information of the sentiment (aspect boundary) label to correct the wrong aspect boundary (sentiment) label. Moreover, we explore two ways of using label embeddings: add and gate mechanism. We evaluate our model on three benchmark datasets. Experimental results verify that our model achieves state-of-the-art performance compared with several baselines.

Aspect-level sentiment classification aims to distinguish the sentiment polarity of each aspect in a given sentence. It is more complex than text-level sentiment classification in that it is a fine-grained task. Existing methods, which formulate this task as predicting the sentiment polarity of a provided (sentence, aspect) pair, tend to ignore the relationship between the sentiment polarity of aspects. In this paper, we propose a sequence prediction model with a sentiment polarity fusion module which sequentially predicts the sentiment polarity of each aspect within sentence. Besides, we use the temporal attention mechanism to keep track of what has been focused on, which discourages repeated attention to the context words with strong sentiment polarity when predicting the sentiment polarity of different aspects. Experimental results on five benchmarking collections illustrate that our proposed model3 outperforms a range of baseline models by a substantial margin, and further demonstrate that the relationship between the sentiment polarity of aspects is helpful to solve the aspect-level sentiment classification.

The task of Knowledge Base Question Answering (KBQA) is to provide a convenient way for the human to more efficiently and easily answer natural language questions using the substantial and valuable knowledge in the KB. Predicate generation is the most important sub-task of KBQA, which aims to generate the predicate paths from head entity to tail entity. Existing models mostly employ a seq2seq method to handle this task. However, the seq2seq method essentially is a classification model, which needs to know the number of the categories in advance. Meanwhile, KBs often are incompleteness and questions are always unbounded, which would cause the problem that the predicates of the questions would be beyond predefined categories. Obviously, the seq2seq model cannot handle this problem. In this paper, to solve the problem above, we explore to build an scoring module with strong genelization to score the predicates not in predefined categories, and also to improve the performance of the seq2seq method. To bridge the gap, we carefully design a reasoning module to score the predicates through reasonably employing the attention mechanism with memory ability. Simultaneously, in order to improve the generalization of the reasoning module, we try to use multi-task learning to enrich the represent of the reasoning module based on the idea of transfer learning. Massive experiments are conducted on two popular benchmark datasets - SimpleQuestion(SimQ) and WebQuestion(WebQ). The experimental results demonstrate that the proposed relation reasoning framework outperforms the state-of-the-art methods.

In recent years, researchers have proposed various methods to achieve better results on the event extraction task. Among them, the pipelined-based methods first perform event trigger prediction and then identify arguments in separate stages. Therefore, the errors from the trigger identification propagate the argument identification. The joint-based methods generally consider the argument role task as a multi-class classification problem, which undoubtedly reduces efficiency. To address these two shortcomings, in this paper, we propose an effective neural model for event extraction. Specifically, we use a Named Entity Recognition method to label arguments while extracting event triggers. As a result, the labels contain arguments and their role information. If the arguments are assigned to event triggers, we assume that there are inter-dependencies between them. Inspired by their inter-dependencies, we then assign arguments to event triggers by matching. This matching operation greatly improves efficiency as its essence is the binary classification. The experimental results on Chinese Emergency Corpus show that our model is more effective and competitive than baselines.

International Classification of Diseases (ICD) codes provides a hierarchy of diagnostic codes for classifying diseases, which plays an important role in healthcare. Medical coding which assigns a subset of ICD codes to a patient visit is a mandatory process that is crucial for patient care and billing. Manual coding is time-consuming, expensive, and error-prone. In this paper, we present a deep learning based network that utilizes the information of textual descriptions and hierarchical structure of ICD codes to make a code assignment automatically. On one hand, we use the machine comprehension framework to model the relationship between the diagnostic description and the textual descriptions, enabling the model put more attention on most relevant segments of diagnostic description for each possible codes. On the other hand, we exploit tree-LSTM architecture to model hierarchical structure, capturing the hierarchical relationship among codes and the semantics of each code. We demonstrate the effectiveness of the proposed methods on a Chinese dataset and an English dataset.

Construction of automatic Question Answering (QA) system for online healthcare community gains great attention from the researchers due to the lack of practitioners who can respond precise answers to patients' consultations. In this paper, we focus on drug treatment question classification task, which can help building QA system for such questions. Due to the lack of labeled data and the high cost of labeling work, we consider using documents or texts from the internet which are related to our task. We fetch a large amount of unlabeled question-answer pairs and drug instructions from the internet and design a co-training style method (which is a semi-supervised learning method) to utilize them. Using some specified models as classifiers, we prove that the classifier trained under our method outperforms the same model trained under supervised learning with less labeled data as input.

Short text clustering is used in various applications and becomes a significant problem, while it also is a challenging task due to the sparsity problem of traditional short text representations. Early methods either cause waste of space or ignore the order of word sequence. To tackle these problems, a self-taught convolutional neural network model is proposed to construct short text representations. However, it extracts the semantic information only from the word context without any other unsupervised features and ignores the different contributions of textual content in clustering. In this paper, we propose an effective short text embedding method for clustering based on word and topic semantic information (STE-WT). Taking advantage of the topic semantic information and capturing the differences in the contributions of the content by an attention mechanism, our proposed model successfully constructs much better short text representations for clustering. Extensive experimental results on real datasets demonstrate the effectiveness and superiority of our framework compared with state-of-the-art methods.

Aspect-based sentiment analysis is a fine-grained task of sentiment classification for multiple aspects in a sentence. While the overall polarity of a sentence has been successfully predicted with neural network architectures, aspect-specific sentiment analysis still faces challenges. In this paper, we propose a novel neural network named IRP-BERT. It is based on BERT which is the newest breakthrough in natural language processing. IRP-BERT aims at improving the accuracy of predicting the right polarity of aspects in a sentence, especially the neutral ones, because some sentimental-related words such as "love" in a noun phrase which is neutrally sentimental could have a wrong impact on the polarity classification to aspects. More specifically, we design a module named Irrelevance-Removing Process (IRP) to counteract the impact from irrelevant sentimental-related features. The final prediction depends on word representations from the outputs of both IRP and BERT. Experiments on a twitter dataset show that our model outperforms both existing state-of-the-art methods and fine-tuned BERT baselines.

How to generate relevant and informative response is one of the core topics in response generation area. Following the task formulation of neural machine translation, previous works mainly consider response generation task as a mapping from a source sentence to a target sentence. However, the dialogue model tends to generate safe, commonplace responses (e.g., I don't know) regardless of the input, when learning to maximize the likelihood of response for the given message in an almost loss-less manner just like MT. Different from existing works, we propose a Global-Local Selective Encoding model (GLSE) to extend the seq2seq framework to generate more relevant and informative responses. Specifically, two types of selective gate network are introduced in this work: (i) A local selective word-sentence gate is added after encoding phase of Seq2Seq learning framework, which learns to tailor the original message information and generates a selected input representation. (ii) A global selective bidirectional-context gate is set to control the bidirectional information flow from a BiGRU based encoder to decoder. Empirical studies indicate the advantage of our model over several classical and strong baselines.

Abstractive summarization is a important task in natural language processing field. In previous work, the sequence-to-sequence based models are widely used for abstractive summarization task. However, most of the current abstractive summarization models still suffer from two problems. One is that it is difficult for these models to learn an accurate source contextual representation from the redundancy and noisy source text at each decoding step. Another is the information loss problem, which is ignored in previous work. The inability of these models to effectively exploit previously generated words led to this problem. In order to address these two problems, in this paper, we propose a novel keyword and generated word attention model. Specifically, the proposed model first employs the hidden state of decoder to capture relevant keywords and previously generated words contextual at each time step. The model then utilizes obtained keywords and generated words contextual to create keywords-aware and generated words-aware source contextual, respectively. The keywords contextual contributes to learn an accurate source contextual representation, and the generated words contextual can alleviate the information loss problem. Experimental results on a popular Chinese social media dataset demonstrate that the proposed model outperforms baselines and achieves the state-of-the-art performance.

Automatic paraphrase generation plays a key role in many natural language applications. The dominant paraphrase generation models are the encoder-decoder neural networks with attention, where the decoder uses the information of the source text while predicting target text. However, the outputs of these paraphrase models often suffer the semantic error problem. This problem is caused by the inadequate information of the decoder. In this work, we introduce a novel neural model to solve this problem, called Collaboration between the Forward and the Backward Decoder. Specifically, the hidden states of the backward decoder are used as supplementary information of the forward decoder. Therefore, the forward decoder can generate more reasonable paraphrase text using the target-side future contextual. Conversely, the backward decoder employs the hidden states of the forward decoder to prevent the semantic error problem. As two experimental examples show, the proposed model can generate the high-quality paraphrase through this collaboration mechanism. The empirical study on two benchmark datasets demonstrates that our model outperforms some baselines and achieves the state-of-the-art performance.

Text summarization and sentiment classification, in NLP, are two main tasks implemented on text analysis, focusing on extracting the major idea of a text at different levels. Based on the characteristics of both, sentiment classification can be regarded as a more abstractive summarization task. According to the scheme, a Self-Attentive Hierarchical model for jointly improving text Summarization and Sentiment Classification (SAHSSC) is proposed in this paper. This model jointly performs abstractive text summarization and sentiment classification within a hierarchical end-to-end neural framework, in which the sentiment classification layer on top of the summarization layer predicts the sentiment label in the light of the text and the generated summary. Furthermore, a self-attention layer is also proposed in the hierarchical framework, which is the bridge that connects the summarization layer and the sentiment classification layer and aims at capturing emotional information at text-level as well as summary-level. The proposed model can generate a more relevant summary and lead to a more accurate summary-aware sentiment prediction. Experimental results evaluated on SNAP amazon online review datasets show that our model outperforms the state-of-the-art baselines on both abstractive text summarization and sentiment classification by a considerable margin.

Origindestination (OD) trip matrices reflect demand patterns of traffic networks and play important roles in traffic engineering. Existing approaches for traffic flow forecast such as ARIMA, SVR, and neural network perform well in modeling and predicting each OD pair separately, but they cause inefficiency when dealing with multidimensional OD demand data. In this paper, to solve the problemgiven OD trip matrices for T moments, how can we predict the overall OD trip matrices at time T+1, T+2, or even T+L, we model temporal vehicle-based OD trip matrix as a four-order tensor consisting of four attributes: origin, destination, vehicle type and time. By resorting to CANDECOMP/PARAFAC (CP) tensor decomposition, we show how a prediction method can be used to forecast future traffic demand in time factor matrix. Experiments based on a real highway tolling dataset demonstrate that our method effectively reduces the time for prediction and meanwhile makes the prediction accuracy competitive with those of other methods.

In partial duplicate image retrieval systems, min-Hash algorithms are widely used because of its high efficiency and robustness. In most of min-Hash algorithms, min-Hash functions are considered independent and grouped into tuples called sketches, the discriminative power of sketches are limited. By modeling correlations of min-Hash functions, we propose a novel sketch construction method called Nonpara-metric Clustering min-Hash (NCmH). In NCmH, the randomly generated min-Hash functions are clustered before grouping them into sketches, while spatial information is fully used in this process. The constructed sketches preserve abundant spatial information between visual words, thus NCmH achieves higher retrieval accuracy compared to the standard min-Hash. Furthermore, our method can be combined with other min-Hash algorithms such as GVP mH [1], PmH [2] and TmH [3] to further improve accuracy. In experiments, we show that our method outperforms the standard min-Hash and improves the state-of-the-art min-Hash algorithm on Oxford 5K dataset and University of Kentucky dataset.

State-of-the-art learning algorithms accept data in feature vector format as input. Examples belonging to different classes may not always be easy to separate in the original feature space. One may ask: can transformation of existing features into new space reveal significant discriminative information not obvious in the original space? Since there can be infinite number of ways to extend features, it is impractical to first enumerate and then perform feature selection. Second, evaluation of discriminative power on the complete dataset is not always optimal. This is because features highly discriminative on subset of examples may not necessarily be significant when evaluated on the entire dataset. Third, feature construction ought to be automated and general, such that, it doesn't require domain knowledge and its improved accuracy maintains over a large number of classification algorithms. In this paper, we propose a framework to address these problems through the following steps: (1) divide-conquer to avoid exhaustive enumeration; (2) local feature construction and evaluation within subspaces of examples where local error is still high and constructed features thus far still do not predict well; (3) weighting rules based search that is domain knowledge free and has provable performance guarantee. Empirical studies indicate that significant improvement (as much as 9% in accuracy and 28% in AUC) is achieved using the newly constructed features over a variety of inductive learners evaluated against a number of balanced, skewed and high-dimensional datasets. Software and datasets are available from the authors.

Traditional machine learning algorithms assume that data are exact or precise. However, this assumption may not hold in some situations because of data uncertainty arising from measurement errors, data staleness, and repeated measurements, etc. With uncertainty, the value of each data item is represented by a probability distribution function (pdf). In this paper, we propose a novel naive Bayes classification algorithm for uncertain data with a pdf. Our key solution is to extend the class conditional probability estimation in the Bayes model to handle pdf’s. Extensive experiments on UCI datasets show that the accuracy of naive Bayes model can be improved by taking into account the uncertainty information.

When labeled examples are limited and difficult to obtain, transfer learning employs knowledge from a source domain to improve learning accuracy in the target domain. However, the assumption made by existing approaches, that the marginal and conditional probabilities are directly related between source and target domains, has limited applicability in either the original space or its linear transformations. To solve this problem, we propose an adaptive kernel approach that maps the marginal distribution of targetdomain and source-domain data into a common kernel space, and utilize a sample selection strategy to draw conditional probabilities between the two domains closer. We formally show that under the kernel-mapping space, the difference in distributions between the two domains is bounded; and the prediction error of the proposed approach can also be bounded. Experimental results demonstrate that the proposed method outperforms both traditional inductive classifiers and the state-of-the-art boosting-based transfer algorithms on most domains, including text categorization and web page ratings. In particular, it can achieve around 10 % higher accuracy than other approaches for the text categorization problem. The source code and datasets are available from the authors.

Transferring knowledge from one domain to another is challenging due to a number of reasons. Since both conditional and marginal distribution of the training data and test data are non-identical, model trained in one domain, when directly applied to a different domain, is usually low in accuracy. For many applications with large feature sets, such as text document, sequence data, medical data, image data of different resolutions, etc. two domains usually do not contain exactly the same features, thus introducing large numbers of "missing values" when considered over the union of features from both domains. In other words, its marginal distributions are at most overlapping. In the same time, these problems are usually high dimensional, such as, several thousands of features. Thus, the combination of high dimensionality and missing values make the relationship in conditional probabilities between two domains hard to measure and model. To address these challenges, we propose a framework that first brings the marginal distributions of two domains closer by "filling up" those missing values of disjoint features. Afterwards, it looks for those comparable sub-structures in the "latent-space" as mapped from the expanded feature vector, where both marginal and conditional distribution are similar. With these sub-structures in latent space, the proposed approach then find common concepts that are transferable across domains with high probability. During prediction, unlabeled instances are treated as "queries", the mostly related labeled instances from out-domain are retrieved, and the classification is made by weighted voting using retrieved out-domain examples. We formally show that importing feature values across domains and latent semantic index can jointly make the distributions of two related domains easier to measure than in original feature space, the nearest neighbor method employed to retrieve related out domain examples is bounded in error when predicting in-domain examples. Software and datasets are available for download.

When the number of labeled examples is limited, traditional supervised feature selection techniques often fail due to sample selection bias or unrepresentative sample problem. To solve this, semi-supervised feature selection techniques exploit the statistical information of both labeled and unlabeled examples in the same time. However, the results of semi-supervised feature selection can be at times unsatisfactory, and the culprit is on how to effectively use the unlabeled data. Quite different from both supervised and semi-supervised feature selection, we propose a ldquohybridrdquoframework based on graph models. We first apply supervised methods to select a small set of most critical features from the labeled data. Importantly, these initial features might otherwise be missed when selection is performed on the labeled and unlabeled examples simultaneously. Next,this initial feature set is expanded and corrected with the use of unlabeled data. We formally analyze why the expected performance of the hybrid framework is better than both supervised and semi-supervised feature selection. Experimental results demonstrate that the proposed method outperforms both traditional supervised and state-of-the-art semi-supervised feature selection algorithms by at least 10% inaccuracy on a number of text and biomedical problems with thousands of features to choose from. Software and dataset is available from the authors.

Sample selection bias is a common problem in many real world applications, where training data are obtained under realistic constraints that make them follow a different distribution from the future testing data. For example, in the application of hospital clinical studies, it is common practice to build models from the eligible volunteers as the training data, and then apply the model to the entire populations. Because these volunteers are usually not selected at random, the training set may not be drawn from the same distribution as the test set. Thus, such a dataset suffers from “sample selection bias” or “covariate shift”. In the past few years, much work has been proposed to reduce sample selection bias, mainly by statically matching the distribution between training set and test set. But in this paper, we do not explore the different distributions directly. Instead, we propose to discover the natural structure of the target distribution, by which different types of sample selection biases can be evidently observed and then be reduced by generating a new sample set from the structure. In particular, unlabeled data are involved in the new sample set to enhance the ability to minimize sample selection bias. One main advantage of the proposed approach is that it can correct all types of sample selection biases, while most of the previously proposed approaches are designed for some specific types of biases. In experimental studies, we simulate all 3 types of sample selection biases on 17 different classification problems, thus 17 × 3 biased datasets are used to test the performance of the proposed algorithm. The baseline models include decision tree, naive Bayes, nearest neighbor, and logistic regression. Across all combinations, the increase in accuracy over non-corrected sample set is 30% on average using each baseline model.