The learnware paradigm proposed by Zhou [34] aims to solve machine learning tasks by leveraging numerous existing high-performing models instead of training from scratch. These models are accommodated in a learnware dock system, where each learnware consists of a model and a specification that characterizes the model's utility, enabling it to be identified for future tasks. A critical challenge in this paradigm remains unresolved: determining what models can be or should be admitted to the system. Without well-established admission criteria, the uncontrolled growth of uploaded models could lead to significant redundancy and inefficiency, resulting in higher storage overhead, increased computational costs, and even potential system failure. To address this gap, this paper presents the first attempt to establish learnware admission criteria and dynamically filter redundant learnwares based on model capability coverage. Specifically, we organize task information from all learnwares into a tree-based structure to assess model capabilities across a continuously expanding task set. Using this structure and model capability representation, we develop an efficient and scalable method for detecting redundant learnwares dynamically without traversing the entire system. Theoretical analysis and extensive experiments involving over ten thousand simulated learnwares validate the efficacy and efficiency of our approach.

We focus on the dueling bandits problem, which has recently drawn significant attention due to its wide-ranging applications in online recommendation systems and the alignment of large language models (LLMs), considers an online preference learning scenario where the learner iteratively selects arms based on pairwise comparison feedback to infer user preferences. Two primary objectives are typically considered in dueling bandits: Regret Minimization (RM), which aims to improve the overall quality of selected arms over time, and Best Arm Identification (BAI), which seeks to efficiently identify the best item with minimal user feedback. For instance, RM is exemplified by the objective of consistently providing high-quality items, while BAI reduces the required human feedback by minimizing the number of necessary comparisons. Conventional research treats RM and BAI as two conflicting objectives, optimizing one at the expense of the other. In this paper, we propose a novel framework that demonstrates the near-consistency of RM and BAI in dueling bandits by reducing the BAI in dueling bandits into a sequential noisy identification problem. Based on our formulation, we propose a black-box reduction technique that transforms any RM algorithm into a BAI algorithm, and prove that such reduction with optimal RM algorithm achieves optimal sample complexity and nearly-optimal cumulative weak regret simultaneously. Our proposed algorithm acheives a nearly-optimal BAI sample complexity and attains a cumulative weak regret that is order-wise equivalent to the best-known result simultaneously. Experiments on both synthetic benchmarks and real-world online recommendation tasks validate the effectiveness of the proposed method, providing empirical evidences for our theoretical findings.

In many real-world applications, data are continuously accumulated in open environments, and new classes may emerge over time. For instance, in disease diagnosis, the prevalence of a certain disease may vary seasonally, and new diseases can also emerge. This paper investigates the problem of learning from unlabeled data stream where thelabel distribution evolves over time, and meanwhile,previously unseen new classes may appear. To handle the emerging new classes in online label shift, we first design a novel risk estimator by unbiased risk rewriting and mixture proportion estimation, which enables the identification of new class data. Subsequently, we employ the online ensemble paradigm for model updating to handle unknown distribution shifts. Moreover, we introduce the sketching and ensemble pruning mechanisms to improve the efficiency of the algorithm, making it more lightweight and practical. The proposed approach enjoys a theoretical guarantee of dynamic regret, ensuring its effectiveness in adapting to the unknown distribution shifts and the emergence of new classes in streaming data. Experiments on diverse benchmark datasets and two real-world applications demonstrate the effectiveness of the algorithm.

Conventional machine learning typically assume a fixed learning objective throughout the learning process.However, for real-world tasks in open and dynamic environments, objectives can change frequently.For example, in autonomous driving, a car has several default modes, but a user's concern for speed and fuel consumption varies depending on road conditions and personal needs.We formulate this problem as learning with varied objectives (LVO), where the goal is to optimize a dynamic weighted combination of multiple sub-objectives by sequentially selecting actions that incur different losses on these sub-objectives.We propose the VaRons algorithm, which estimates the action-wise performance on each sub-objective and adaptively selects decisions according to the dynamic requirements on different sub-objectives.Further, we extend our approach to cases involving contextual representations and propose the Con-VaRons algorithm, assuming parameterized linear structure that links contextual features to the main objective.Both the VaRons and ConVaRons are provably minimax optimal with respect to the time horizon 𝑇 , with ConVaRons showing better dependency with the number of sub-objectives 𝐾.Experiments on dynamic classifier and real-world cluster service allocation tasks validate the effectiveness of our methods and support our theoretical findings.

The learnware paradigm proposed by Zhou [40] aims to enable users to leverage numerous existing high-performing models instead of building machine learning models from scratch.This paradigm envisions that: Any developer worldwide can submit their well-trained models spontaneously into a learnware dock system (formerly known as learnware market).The system uniformly generates a specification for each model to form a learnware and accommodates it.As the key component, a specification should represent the capabilities of the model while preserving developer's original data.Based on the specifications, the learnware dock system can identify and assemble existing learnwares for users to solve new machine learning tasks.Recently, based on reduced kernel mean embedding (RKME) specification, a series of studies have shown the effectiveness of the learnware paradigm theoretically and empirically.However, the realization of a learnware dock system is still missing and remains a big challenge.This paper proposes Beimingwu, the first open-source learnware dock system, providing foundational support for future research.The system provides implementations and extensibility for the entire process of learnware paradigm, including the submitting, usability testing, organization, identification, deployment, and reuse of learnwares.Utilizing Beimingwu, the model development for new user tasks can be significantly streamlined, thanks to integrated architecture and engine design, specifying unified learnware structure and scalable APIs, and the integration of various algorithms for learnware identification and reuse.Notably, this is possible even for users with limited data and minimal expertise in machine learning, without compromising the raw data's security.The system facilitates the future research implementations in learnware-related algorithms and systems, and lays the ground for hosting a vast array of learnwares and establishing a learnware ecosystem.The system is fully open-source and we expect the research community

Learning with data streams has attracted much attention in recent decades. Conventional approaches typically assume that the feature and label of a data item can be timely observed at each round. In many real-world tasks, however, it often occurs that either the feature or the label is observed firstly while the other arrives with delay. For instance, in distributed learning systems, a central processor collects training data from different sub-processors to train a learning model, whereas the feature and label of certain data items can arrive asynchronously due to network latency. The problem of learning with asynchronous feature or label in streams encompasses many applications but still lacks sound solutions. In this article, we formulate the problem and propose a new approach to alleviate the negative effect of asynchronicity and mining asynchronous data streams. Our approach carefully exploits the timely arrived information and builds an online ensemble structure to adaptively reuse historical models and instances. We provide the theoretical guarantees of our approach and conduct extensive experiments to validate its effectiveness.

In many real-world applications, data are continuously accumulated within open environments. For instance, in disease diagnosis, the prevalence of diseases can vary across seasons, and new types of diseases can emerge. This paper investigates the problem of learning from unlabeled data where the label distribution evolves over time, and meanwhile, previously unseen new class appears in the data stream. To handle the new class in online label shift, we first design a novel risk estimator by unbiased risk rewriting and mixture proportion estimation. Subsequently, we employ the online ensemble paradigm for model updating to handle unknown distribution shifts. The proposed approach enjoys a theoretical guarantee of dynamic regret, ensuring its effectiveness in adapting to the changing label distribution and the presence of the new class in streams. Experiments conducted on diverse benchmark datasets and two real-world applications demonstrate the effectiveness of the proposed algorithm.

Discovering causal relations from observational data is at the heart of scientific research. Most causal discovery methods assume that the data have only one variable type. In real-world problems, however, data can consist of a mixture of continuous, discrete, and categorical variables. In this paper, we examine the causal discovery problem on mixed data. We introduce a general tree-structured functional causal model, which is well suited for characterizing the generating mechanisms of mixed data by allowing non- differentiability and nonlinearity. We present corresponding identifiability results, showing that under mild conditions, the causal directions can be uniquely determined from observational distributions. Further, we prove that the causal direction between continuous and discrete variables is generally identifiable under a much larger function class. Based on the theoretical findings, we propose an effective causal discovery method leveraging a consistent score function and powerful tree-learning techniques. Experiments on both synthetic and real data verify the effectiveness of our approach.

Most existing LTR approaches follow a supervised learning paradigm from offline data collected from the online system. However, it has been noticed that previous LTR models can have good performances over offline validation data but have poor online performances, which implies a possible large inconsistency between the offline and online evaluation. We investigate and confirm in this paper that such inconsistency exists and can have a significant impact on AliExpress Search. Reasons for the inconsistency include the ignorance of item context. Therefore, this paper proposes an evaluator-generator framework for LTR with item context. The framework consists of an evaluator that generalizes to evaluate recommendations involving the context, and a generator that maximizes the evaluator score by reinforcement learning, and a discriminator that ensures the generalization of the evaluator. Extensive experiments in simulation environments and AliExpress Search online system show that, firstly, the classic data-based metrics on the offline dataset can show significant inconsistency with online performance. Secondly, the proposed evaluator score is significantly more consistent with the online performance than common ranking metrics. Finally, as the consequence, our method achieves a significant improvement in terms of Conversion Rate over the industrial-level fine-tuned model in online A/B tests.

We introduce Isolation Distributional Kernel as a new way to measure the similarity between two distributions. Existing approaches based on kernel mean embedding, which convert a point kernel to a distributional kernel, have two key issues: the point kernel employed has a feature map with intractable dimensionality; and it is {\em data independent}. This paper shows that Isolation Distributional Kernel (IDK), which is based on a {\em data dependent} point kernel, addresses both key issues. We demonstrate IDK's efficacy and efficiency as a new tool for kernel based anomaly detection for both point and group anomalies. Without explicit learning, using IDK alone outperforms existing kernel based point anomaly detector OCSVM and other kernel mean embedding methods that rely on Gaussian kernel. For group anomaly detection,we introduce an IDK based detector called IDK$^2$. It reformulates the problem of group anomaly detection in input space into the problem of point anomaly detection in Hilbert space, without the need for learning. IDK$^2$ runs orders of magnitude faster than group anomaly detector OCSMM.We reveal for the first time that an effective kernel based anomaly detector based on kernel mean embedding must employ a characteristic kernel which is data dependent.

Given a publicly available pool of machine learning models constructed for various tasks, when a user plans to build a model for her own machine learning application, is it possible to build upon models in the pool such that the previous efforts on these existing models can be reused rather than starting from scratch? Here, a grand challenge is how to find models that are helpful for the current application, without accessing the raw training data for the models in the pool. In this paper, we present a two-phase framework. In the upload phase, when a model is uploading into the pool, we construct a reduced kernel mean embedding (RKME) as aspecificationfor the model. Then in the deployment phase, the relatedness of the current task and pre-trained models will be measured based on the value of the RKME specification. Theoretical results and extensive experiments validate the effectiveness of our approach.

In plenty of real-world applications, data are collected in a streaming fashion, and their accurate labels are hard to obtain. For instance, in the environmental monitoring task, sensors are collecting the data all the time. Still, their labels are scarce because the labeling process requires human effort and can conceal annotation errors. This paper investigates the problem of learning with weakly labeled data streams, in which data are continuously collected, and only a limited subset of streaming data is labeled but potentially with noise. This setting is challenging and of great importance but rarely studied in the literature. When the data are constantly gathered with unknown noise on labels, it is quite challenging to design algorithms to obtain a well-generalized classifier. To address this difficulty, we propose a novel noise transition matrix estimation approach for data streams with scarce noisy labels by online anchor points identification. Based on that, we propose an adaptive learning algorithm for weakly labeled data streams via model reuse and effectively alleviate the negative influence of label noise with unlabeled data. Both theoretical analysis and extensive experiments justify and validate the effectiveness of the proposed approach.

Most studies about deep learning are based on neural network models, where many layers of parameterized nonlinear differentiable modules are trained by backpropagation. Recently, it has been shown that deep learning can also be realized by non-differentiable modules without backpropagation training called deep forest. We identify that deep forest has high time costs and memory requirements—this has inhibited its use on large-scale datasets. In this paper, we propose a simple and effective approach with three main strategies for efficient learning of deep forest. First, it substantially reduces the number of instances that needs to be processed through redirecting instances having high predictive confidence straight to the final level for prediction, by-passing all the intermediate levels. Second, many non-informative features are screened out, and only the informative ones are used for learning at each level. Third, an unsupervised feature transformation procedure is proposed to replace the supervised multi-grained scanning procedure. Our theoretical analysis supports the proposed approach in varying the model complexity from low to high as the number of levels increases in deep forest. Experiments show that our approach achieves highly competitive predictive performance with reduced time cost and memory requirement by one to two orders of magnitude.

In plenty of real-life tasks, strongly supervised information is hard to obtain, and thus weakly supervised learning has drawn considerable attention recently. This paper investigates the problem of learning from incomplete and inaccurate supervision, where only a limited subset of training data is labeled but potentially with noise. This setting is challenging and of great importance but rarely studied in the literature. We notice that in many applications, the limited labeled data are with certain structures, which paves us a way to design effective methods. Specifically, we observe that labeled data are usually with one-sided noise such as the bug detection task, where the identified buggy codes are indeed with defects, while codes checked many times or newly fixed may still have other flaws. Furthermore, when there occurs two-sided noise in the labeled data, we exploit the class-prior information of unlabeled data, which is typically available in practical tasks. We propose novel approaches for the incomplete and inaccurate supervision learning tasks and effectively alleviate the negative influence of label noise with the help of a vast number of unlabeled data. Both theoretical analysis and extensive experiments justify and validate the effectiveness of the proposed approaches.

Multivariate time series (MTS) prediction is ubiquitous in real-world fields, but MTS data often contains missing values. In recent years, there has been an increasing interest in using end-to-end models to handle MTS with missing values. To generate features for prediction, existing methods either merge all input dimensions of MTS or tackle each input dimension independently. However, both approaches are hard to perform well because the former usually produce many unreliable features and the latter lacks correlated information. In this paper, we propose a Learning Individual Features (LIFE) framework, which provides a new paradigm for MTS prediction with missing values. LIFE generates reliable features for prediction by using the correlated dimensions as auxiliary information and suppressing the interference from uncorrelated dimensions with missing values. Experiments on two real-world data sets verify the superiority of LIFE to existing state-of-the-art models. The full version of this work can refer to arXiv (2109.14844).

Stroke evaluation is critical for coaches to evaluate players' performance in table tennis matches. However, current methods highly demand proficient knowledge in table tennis and are time-consuming. We collaborate with the Chinese national table tennis team and propose Tac-Valuer, an automatic stroke evaluation framework for analysts in table tennis teams. In particular, to integrate analysts' knowledge into the machine learning model, we employ the latest effective framework named abductive learning, showing promising performance. Based on abductive learning, Tac-Valuer combines the state-of-the-art computer vision algorithms to extract and embed stroke features for evaluation. We evaluate the design choices of the approach and present Tac-Valuer's usability through use cases that analyze the performance of the top table tennis players in world-class events.

Learning with streaming data has attracted much attention during the past few years. Though most studies consider data stream with fixed features, in real practice the features may be evolvable. For example, features of data gathered by limited-lifespan sensors will change when these sensors are substituted by new ones. In this article, we propose a novel learning paradigm: Feature Evolvable Streaming Learning where old features would vanish and new features would occur. Rather than relying on only the current features, we attempt to recover the vanished features and exploit it to improve performance. Specifically, we learn a mapping from the overlapping period to recover old features and then we learn two models from the recovered features and the current features, respectively. To benefit from the recovered features, we develop two ensemble methods. In the first method, we combine the predictions from two models and theoretically show that with the assistance of old features, the performance on new features can be improved and we provide a tighter bound when the loss function is exponentially concave. In the second approach, we dynamically select the best single prediction and establish a better performance guarantee when the best model switches. Experiments on both synthetic and real data validate the effectiveness of our proposal.

Diverse applications involving complicated data objects such as proteins and images are solved by applying multi-instance learning (MIL) algorithms. However, few MIL algorithms can deal with problems in an open and dynamic environment, where new categories of samples emerge. In this type of emerging novel class setting, algorithms should be able to not only classify the samples from the observed classes accurately, but also recognize the samples from the novel class. In this paper, we focus on the Multi-Instance learning with Emerging Novel class (MIEN) problem, and formulate MIEN from a metric learning perspective. We extract key instances to form the “super-bag” for each observed class, and non-key instances from all the observed classes to form a “meta super-bag”. Based on these super-bags, we propose the MIEN-metric method to learn discriminative metrics for classifying MIL bags from the observed classes and recognizing bags from the novel class. Experimental results of diverse domains, e.g., biological function annotation, text categorization, and object-centric/scene-centric image classification, show MIEN-metric outperforms other baseline methods significantly when the novel class emerges. Meanwhile, MIEN-metric is comparable with state-of-the-art MIL algorithms for binary classification in the traditional MIL setting.

In many large-scale machine learning applications, data are accumulated over time, and thus, an appropriate model should be able to update in an online style. In particular, it would be ideal to have a storage independent from the data volume, and scan each data item only once. Meanwhile, the data distribution usually changes during the accumulation procedure, making distribution-free one-pass learning a challenging task. In this paper, we propose a simple yet effective approach for this task, without requiring prior knowledge about the change, where every data item can be discarded once scanned. We also present a variant for high-dimensional situations, by exploiting compressed sensing to reduce computational and storage complexity. Theoretical analysis shows that our proposal converges under mild assumptions, and the performance is validated on both synthetic and real-world datasets.

In many practical tasks, there are usually two kinds of common information: cheap unlabeled data and domain knowledge in the form of symbols. There are some attempts using one single information source, such as semi-supervised learning and abductive learning. However, there is little work to use these two kinds of information sources at the same time, because it is very difficult to combine symbolic logical representation and numerical model optimization effectively. The learning becomes even more challenging when the domain knowledge is insufficient. In this paper, we present an attempt-Semi-Supervised ABductive Learning (SS-ABL) framework. In this framework, semi-supervised learning is trained via pseudo labels of unlabeled data generated by abductive learning, and the background knowledge is refined via the label distribution predicted by semi-supervised learning. The above framework can be optimized iteratively and can be naturally interpretable. The effectiveness of our framework has been fully verified in the theft judicial sentencing of real legal documents. In the case of missing sentencing elements and mixed legal rules, our framework is apparently superior to many existing baseline practices, and provides explanatory assistance to judicial sentencing.

We introduce Isolation Distributional Kernel as a new way to measure the similarity between two distributions. Existing approaches based on kernel mean embedding, which converts a point kernel to a distributional kernel, have two key issues: the point kernel employed has a feature map with intractable dimensionality; and it is data independent. This paper shows that Isolation Distributional Kernel (IDK), which is based on a data dependent point kernel, addresses both key issues. We demonstrate IDK's efficacy and efficiency as a new tool for kernel based anomaly detection. Without explicit learning, using IDK alone outperforms existing kernel based anomaly detector OCSVM and other kernel mean embedding methods that rely on Gaussian kernel. We reveal for the first time that an effective kernel based anomaly detector based on kernel mean embedding must employ a characteristic kernel which is data dependent.

The word "deep learning" is generally regarded as a synonym of "deep neural networks (DNNs)". In this talk, we will discuss on essentials in deep learning and claim that deep learning is not necessarily to be realized by neural networks and differentiable modules. We will then present an exploration to non-NN style deep learning, where the building blocks are non-differentiable modules and the training process does not rely on backpropagation or gradient-based adjustment. We will also talk about some recent advances and challenges in this direction of research.

Multi-label learning methods assign multiple labels to one object. In practice, in addition to differentiating relevant labels from irrelevant ones, it is often desired to rank relevant labels for an object, whereas the ranking of irrelevant labels is not important. Thus, we require an algorithm to do classification and ranking of relevant labels simultaneously. Such a requirement, however, cannot be met because most existing methods were designed to optimize existing criteria, yet there is no criterion which encodes the aforementioned requirement. In this paper, we present a new criterion, PRO LOSS, concerning the prediction of all labels as well as the ranking of only relevant labels. We then propose ProSVM which optimizes PRO LOSS efficiently using alternating direction method of multipliers. We further improve its efficiency with an upper approximation that reduces the number of constraints from O(T2) to O(T), where T is the number of labels. We then notice that in real applications, it is difficult to get full supervised information for multi-label data. To make the proposed algorithm more robust to supervised information, we adapt ProSVM to deal with the multi-label learning with partial labels problem. Experiments show that our proposal is not only superior on PRO LOSS, but also highly competitive on existing evaluation criteria.

Support Vector Machine (SVM) has always been one of the most successful learning algorithms, with the central idea of maximizing theminimum margin, i.e., the smallest distance from the instances to the classification boundary. However, recent theoretical results disclosed that maximizing the minimum margin does not necessarily lead to better generalization performance, and instead, themargin distributionhas been proven to be more crucial. Based on this idea, we propose the Optimal margin Distribution Machine (ODM), which can achieve a better generalization performance by optimizing the margin distribution explicitly. We characterize the margin distribution by the first- and second-order statistics, i.e., the margin mean and variance. The proposed method is a general learning approach which can be applied in any place where SVMs are used, and its superiority is verified both theoretically and empirically in this paper.

Set-level problems are as important as instance-level problems. The core in solving set-level problems is: how to measure the similarity between two sets. This paper investigates data-dependent kernels that are derived directly from data. We introduce Isolation Set-Kernel which is solely dependent on data distribution, requiring neither class information nor explicit learning. In contrast, most current set-similarities are not dependent on the underlying data distribution. We theoretically analyze the characteristic of Isolation Set-Kernel. As the set-kernel has a finite feature map, we show that it can be used to speed up the set-kernel computation significantly. We apply Isolation Set-Kernel to Multi-Instance Learning (MIL) using SVM classifier, and demonstrate that it outperforms other set-kernels or other solutions to the MIL problem.

In plenty of real-life tasks, strongly supervised information is hard to obtain, such that there is not sufficient high-quality supervision to make traditional learning approaches succeed. Therefore, weakly supervised learning has drawn considerable attention recently. In this paper, we consider the problem of learning from incomplete and inaccurate supervision, where only a limited subset of training data is labeled but potentially with noise. This setting is challenging and of great importance but rarely studied in the literature. We notice that in many applications, the limited labeled data are usually with one-sided noise. For instance, considering the bug detection task in the software system, the identified buggy codes are indeed with defects whereas the codes that have been checked many times or newly fixed may still have other flaws due to the complexity of the system. We propose a novel method which is able to effectively alleviate the negative influence of one-sided label noise with the help of a vast number of unlabeled data. Excess risk analysis is provided as theoretical justifications on the usefulness of incomplete and one-sided inaccurate supervision. We conduct experiments on synthetic, benchmark datasets, and real-life tasks to validate the effectiveness of the proposed approach.

Multi-Instance Multi-Label (MIML) learning is a popular framework in machine learning, where each object is represented by a bag of instances, and associated with multiple labels. While MIML learning has achieved success in many applications, it is less clear how the labels are related to the instances. In this paper, we propose to study the causal relation between instances and labels, which on one hand can improve the interpretability of complicated MIML models, and on the other hand may further improve the prediction performance at both instance and bag levels. We exploit prototypes in the instance space as a bridge to represent the examples, and then propose an efficient algorithm to identify the causal relations from prototypes to class labels, which are further utilized for model training and key instance detection. Experiments on various datasets show that in addition to superior classification performance, our approach can identify reasonable causal relations between instances and labels.

Internet companies are facing the need for handling large-scale machine learning applications on a daily basis and distributed implementation of machine learning algorithms which can handle extra-large-scale tasks with great performance is widely needed. Deep forest is a recently proposed deep learning framework which uses tree ensembles as its building blocks and it has achieved highly competitive results on various domains of tasks. However, it has not been tested on extremely large-scale tasks. In this work, based on our parameter server system, we developed the distributed version of deep forest. To meet the need for real-world tasks, many improvements are introduced to the original deep forest model, including MART (Multiple Additive Regression Tree) as base learners for efficiency and effectiveness consideration, the cost-based method for handling prevalent class-imbalanced data, MART based feature selection for high dimension data, and different evaluation metrics for automatically determining the cascade level. We tested the deep forest model on an extra-large-scale task, i.e., automatic detection of cash-out fraud, with more than 100 million training samples. Experimental results showed that the deep forest model has the best performance according to the evaluation metrics from different perspectives even with very little effort for parameter tuning. This model can block fraud transactions in a large amount of money each day. Even compared with the best-deployed model, the deep forest model can additionally bring a significant decrease in economic loss each day.

We consider multi-label crowdsourcing learning in two scenarios. In the first scenario, we aim at inferring instances' groundtruth given the crowds' annotations. We propose two approaches NAM/RAM (Neighborhood/Relevance Aware Multi-label crowdsourcing) modeling the crowds' expertise and label correlations from different perspectives. Extended from single-label crowdsourcing methods, NAM models the crowds' expertise on individual labels, but based on the idea that for rational workers, their annotations for instances similar in the feature space should also be similar, NAM utilizes information from the feature space and incorporates the local influence of neighborhoods' annotations. Noting that the crowds tend to act in an effort-saving manner while labeling multiple labels, i.e., rather than carefully annotating every proper label, they would prefer scanning and tagging a few most relevant labels, RAM models the crowds' expertise as their ability to distinguish the relevance between label pairs. In the second scenario, we care about cost-efficient crowdsourcing where the labeling and learning process are conducted in tandem. We extend NAM/RAM to the active paradigm and propose instance, label, and worker selection criteria such that the labeling cost is significantly saved compared to passive learning without labeling control. The proposals' effectiveness are validated on simulated and real data.

Most studies about deep learning are based on neural network models, where many layers of parameterized nonlinear differentiable modules are trained by backpropagation. Recently, it has been shown that deep learning can also be realized by non-differentiable modules without backpropagation training called deep forest. The developed representation learning process is based on a cascade of cascades of decision tree forests, where the high memory requirement and the high time cost inhibit the training of large models. In this paper, we propose a simple yet effective approach to improve the efficiency of deep forest. The key idea is to pass the instances with high confidence directly to the final stage rather than passing through all the levels. We also provide a theoretical analysis suggesting a means to vary the model complexity from low to high as the level increases in the cascade, which further reduces the memory requirement and time cost. Our experiments show that the proposed approach achieves highly competitive predictive performance with significantly reduced time cost and memory requirement by up to one order of magnitude.

This paper investigates data dependent kernels that are derived directly from data. This has been an outstanding issue for about two decades which hampered the development of kernel-based methods. We introduce Isolation Kernel which is solely dependent on data distribution, requiring neither class information nor explicit learning to be a classifier. In contrast, existing data dependent kernels rely heavily on class information and explicit learning to produce a classifier. We show that Isolation Kernel approximates well to a data independent kernel function called Laplacian kernel under uniform density distribution. With this revelation, Isolation Kernel can be viewed as a data dependent kernel that adapts a data independent kernel to the structure of a dataset. We also provide a reason why the proposed new data dependent kernel enables SVM (which employs a kernel through other means) to improve its predictive accuracy. The key differences between Random Forest kernel and Isolation Kernel are discussed to examine the reasons why the latter is a more successful tree-based kernel.

It is well-known that exploiting label correlations is important to multi-label learning. Existing approaches either assume that the label correlations are global and shared by all instances; or that the label correlations are local and shared only by a data subset. In fact, in the real-world applications, both cases may occur that some label correlations are globally applicable and some are shared only in a local group of instances. Moreover, it is also a usual case that only partial labels are observed, which makes the exploitation of the label correlations much more difficult. That is, it is hard to estimate the label correlations when many labels are absent. In this paper, we propose a new multi-label approach GLOCAL dealing with both the full-label and the missing-label cases, exploiting global and local label correlations simultaneously, through learning a latent label representation and optimizing label manifolds. The extensive experimental studies validate the effectiveness of our approach on both full-label and missing-label data.

In a multi-label learning task, an object possesses multiple concepts where each concept is represented by a class label. Previous studies on multi-label learning have focused on a fixed set of class labels, i.e., the class label set of test data is the same as that in the training set. In many applications, however, the environment is dynamic and new concepts may emerge in a data stream. In order to maintain a good predictive performance in this environment, a multi-label learning method must have the ability to detect and classify instances with emerging new labels. To this end, we propose a new approach called Multi-label learning with Emerging New Labels (MuENL). It has three functions: classify instances on currently known labels, detect the emergence of a new label, and construct a new classifier for each new label that works collaboratively with the classifier for known labels. In addition, we show that MuENL can be easily extended to handle sparse high dimensional data streams by simply reducing the original dimensionality, and then applying MuENL on the reduced dimensional space. Our empirical evaluation shows the effectiveness of MuENL on several benchmark datasets and MuENLHD on the sparse high dimensional Weibo dataset.

Friendship is the cornerstone to build a social network. In online social networks, statistics show that the leading reason for user to create a new friendship is due to recommendation. Thus the accuracy of recommendation matters. In this paper, we propose a Bayesian Personalized Ranking Deep Neural Network (BayDNN) model for friend recommendation in social networks. With BayDNN, we achieve significant improvement on two public datasets: Epinions and Slashdot. For example, on Epinions dataset, BayDNN significantly outperforms the state-of-the-art algorithms, with a 5% improvement on NDCG over the best baseline.

One pass learning updates a model with only a single scan of the dataset, without storing historical data. Previous studies focus on classification tasks with a fixed class set, and will perform poorly in an open dynamic environment when new classes emerge in a data stream. The performance degrades because the classifier needs to receive a sufficient number of instances from new classes to establish a good model. This can take a long period of time. In order to reduce this period to deal with any-time prediction task, we introduce a framework to handle emerging new classes called One-Pass Class Incremental Learning (OPCIL). The central issue in OPCIL is: how to effectively adapt a classifier of existing classes to incorporate emerging new classes. We call it the new class adaptation issue, and propose a new approach to address it, which requires only one new class instance. The key is to generate pseudo instances which are optimized to satisfy properties that produce a good discriminative classifier. We provide the necessary propertiesand optimization procedures required to address this issue. Experiments validate the effectiveness of this approach.

This paper investigates an important problem in stream mining, i.e., classification under streaming emerging new classes or SENC. The SENC problem can be decomposed into three subproblems: detecting emerging new classes, classifying known classes, and updating models to integrate each new class as part of known classes. The common approach is to treat it as a classification problem and solve it using either a supervised learner or a semi-supervised learner. We propose an alternative approach by using unsupervised learning as the basis to solve this problem. The proposed method employs completely-random trees which have been shown to work well in unsupervised learning and supervised learning independently in the literature. The completely-random trees are used as a single common core to solve all three subproblems: unsupervised learning, supervised learning, and model update on data streams. We show that the proposed unsupervised-learning-focused method often achieves significantly better outcomes than existing classification-focused methods.

Multi-label learning is widely applied in many tasks, where an object possesses multiple concepts with each represented by a class label. Previous studies on multi-label learning have focused on a fixed set of class labels, i.e., the class label set of test data is the same as that in the training set. In many applications, however, the environment is open and new concepts may emerge with previously unseen instances. In order to maintain good predictive performance in this environment, a multi-label learning method must have the ability to detect and classify those instances with emerging new labels. To this end, we propose a new approach called Multi-label learning with Emerging New Labels (MuENL). It builds models with three functions: classify instances on currently known labels, detect the emergence of a new label in new instances, and construct a new classifier for each new label that works collaboratively with the classifier for known labels. Our empirical evaluation shows the effectiveness of MuENL.

User identity linkage across social platforms is an important problem of great research challenge and practical value. In real applications, the task often assumes an extra degree of difficulty by requiring linkage across multiple platforms. While pair-wise user linkage between two platforms, which has been the focus of most existing solutions, provides reasonably convincing linkage, the result depends by nature on the order of platform pairs in execution with no theoretical guarantee on its stability. In this paper, we explore a new concept of ``Latent User Space'' to more naturally model the relationship between the underlying real users and their observed projections onto the varied social platforms, such that the more similar the real users, the closer their profiles in the latent user space. We propose two effective algorithms, a batch model(ULink) and an online model(ULink-On), based on latent user space modelling. Two simple yet effective optimization methods are used for optimizing objective function: the first one based on the constrained concave-convex procedure(CCCP) and the second on accelerated proximal gradient. To our best knowledge, this is the first work to propose a unified framework to address the following two important aspects of the multi-platform user identity linkage problem --- (I) the platform multiplicity and (II) online data generation. We present experimental evaluations on real-world data sets for not only traditional pairwise-platform linkage but also multi-platform linkage. The results demonstrate the superiority of our proposed method over the state-of-the-art ones.

This paper introduces the first generic version of data dependent dissimilarity and shows that it provides a better closest match than distance measures for three existing algorithms in clustering, anomaly detection and multi-label classification. For each algorithm, we show that by simply replacing the distance measure with the data dependent dissimilarity measure, it overcomes a key weakness of the otherwise unchanged algorithm.

It is traditionally a challenge for home buyers to understand, compare, and contrast the investment value of real estate. Although a number of appraisal methods have been developed to value real properties, the performances of these methods have been limited by traditional data sources for real estate appraisal. With the development of new ways of collecting estate-related mobile data, there is a potential to leverage geographic dependencies of real estate for enhancing real estate appraisal. Indeed, the geographic dependencies of the investment value of an estate can be from the characteristics of its own neighborhood (individual), the values of its nearby estates (peer), and the prosperity of the affiliated latent business area (zone). To this end, in this paper, we propose a geographic method, named ClusRanking, for real estate appraisal by leveraging the mutual enforcement of ranking and clustering power. ClusRanking is able to exploit geographic individual, peer, and zone dependencies in a probabilistic ranking model. Specifically, we first extract the geographic utility of estates from geography data, estimate the neighborhood popularity of estates by mining taxicab trajectory data, and model the influence of latent business areas. Also, we fuse these three influential factors and predict real estate investment value. Moreover, we simultaneously consider individual, peer and zone dependencies, and derive an estate-specific ranking likelihood as the objective function. Furthermore, we propose an improved method named CR-ClusRanking by incorporating checkin information as a regularization term which reduces the performance volatility of real estate ranking system. Finally, we conduct a comprehensive evaluation with the real estate-related data of Beijing, and the experimental results demonstrate the effectiveness of our proposed methods.

In many real-world applications, different types of misclassification usually suffer from different costs, but the accurate cost is often hard to be determined and usually one can only get an interval-estimation like that one type of mistake is about 5 to 10 times more serious than the other type. On the other hand, there are usually abundant unlabeled data available, leading to great research effort about semi-supervised learning. It is noticeable that cost interval and unlabeled data usually appear simultaneously in practice tasks; however, there is rare study tackling them together. In this paper, we propose the cisLDM approach which is able to handle cost interval and exploit unlabeled data in a principled way. Rather than maximizing the minimum margin like traditional large margin classifiers, cisLDM tries to optimize the margin distribution on both labeled and unlabeled data when minimizing the worst-case totalcost and the mean total-cost simultaneously according to the cost interval. Experiments on a broad range of datasets and cost settings exhibit the impressive performance of cisLDM. In particular, cisLDM is able to reduce 47 percent more total-cost than standard SVM and 27 percent more total-cost than cost-sensitive semi-supervised SVM which assumes the true cost value is known in advance.

Complex media objects are often described by multi-view feature groups collected from diverse domains or information channels. Multi-view learning, which attempts to exploit the relationship among multiple views to improve learning performance, has drawn extensive attention. It is noteworthy that in some real-world applications, features of different views may come from different private data repositories, and thus, it is desired to exploit view relationship with data privacy preserved simultaneously. Existing multi-view learning approaches such as subspace methods and pre-fusion methods are not applicable in this scenario because they need to access the whole features, whereas late-fusion approaches could not exploit information from other views to improve the individual view-specific learners. In this paper, we propose a novel multi-view learning framework which works in a hybrid fusion manner. Specifically, we convert predicted values of each view into an Accumulated Prediction Matrix (APM) with low-rank constraint enforced jointly by the multiple views. The joint low-rank constraint enables the view-specific learner to exploit other views to help improve the performance, without accessing the features of other views. Thus, the proposed RANC framework provides a privacy-preserving way for multi-view learning. Furthermore, we consider variants of solutions to achieve rank consistency and present corresponding methods for the optimization. Empirical investigations on real datasets show that the proposed method achieves state-of-the-art performance on various tasks.

In real tasks, usually a good classification performance can only be obtained when a good distance metric is obtained; therefore, distance metric learning has attracted significant attention in the past few years. Typical studies of distance metric learning evaluate how to construct an appropriate distance metric that is able to separate training data points from different classes or satisfy a set of constraints (e.g., must-links and/or cannot-links). It is noteworthy that this task becomes challenging when there are only limited labeled training data points and no constraints are given explicitly. Moreover, most existing approaches aim to construct a global distance metric that is applicable to all data points. However, different data points may have different properties and may require different distance metrics. We notice that data points in real tasks are often connected by physical links (e.g., people are linked with each other in social networks; personal webpages are often connected to other webpages, including nonpersonal webpages), but the linkage information has not been exploited in distance metric learning. In this article, we develop a pairwised specific distance (PSD) approach that exploits the structures of physical linkages and in particular captures the key observations that nonmetric and clique linkages imply the appearance of different or unique semantics, respectively. It is noteworthy that, rather than generating a global distance, PSD generates different distances for different pairs of data points; this property is desired in applications involving complicated data semantics. We mainly present PSD for multi-class learning and further extend it to multi-label learning. Experimental results validate the effectiveness of PSD, especially in the scenarios in which there are very limited labeled training data points and no explicit constraints are given.

Automatic location semantics prediction for living analytics based on smartphone data has attracted extensive attention in just recent years. Basically, this task can be formulated as a multi-class classification problem, where different location/places are regarded as different labels. Previous studies were mostly based on common classification techniques directly, neglecting the critical challenging issue of class imbalance in such a problem (e.g., people go to offices much more often than they go to cinemas). It is also noteworthy that in contrast to common multi-class problems where the classes can be treated independently and interchangeably, the places for labeling usually have important correlations, which should be taken account in the classification/labeling process. Moreover, several activities may occur in the same place and thus the same place label might convey different semantics. In this paper, we address the above issues for location semantics prediction by proposing the FS-Mining (Frame-based Semantics Mining) approach. We treat the raw sensor data in the smartphone as a sequence of short and non-overlapping frames, based on which the user behavior at each place can be characterized and the place semantics can be modeled. To deal with the issues of label relation and class imbalance, a multi-level classification model with class-split and class-merge mechanisms was also developed. An ensemble strategy was also employed to further improve the performance. Experiments on the dataset of Nokia Mobile Data Challenge [1] demonstrate promising performances for the FS-Mining approach.

Multi-instance learning (MIL) has been widely applied to diverse applications involving complicated data objects such as images and genes. However, most existing MIL algorithms can only handle small-or moderate-sized data. In order to deal with the large scale problems in MIL, we propose an efficient and scalable MIL algorithm named miFV. Our algorithm maps the original MIL bags into a new feature vector representation, which can obtain bag-level information, and meanwhile lead to excellent performances even with linear classifiers. In consequence, thanks to the low computational cost in the mapping step and the scalability of linear classifiers, miFV can handle large scale MIL data efficiently and effectively. Experiments show that miFV not only achieves comparable accuracy rates with state-of-the-art MIL algorithms, but has hundreds of times faster speed than other MIL algorithms.

It is traditionally a challenge for home buyers to understand, compare and contrast the investment values of real estates. While a number of estate appraisal methods have been developed to value real property, the performances of these methods have been limited by the traditional data sources for estate appraisal. However, with the development of new ways of collecting estate-related mobile data, there is a potential to leverage geographic dependencies of estates for enhancing estate appraisal. Indeed, the geographic dependencies of the value of an estate can be from the characteristics of its own neighborhood (individual), the values of its nearby estates (peer), and the prosperity of the affiliated latent business area (zone). To this end, in this paper, we propose a geographic method, named ClusRanking, for estate appraisal by leveraging the mutual enforcement of ranking and clustering power. ClusRanking is able to exploit geographic individual, peer, and zone dependencies in a probabilistic ranking model. Specifically, we first extract the geographic utility of estates from geography data, estimate the neighborhood popularity of estates by mining taxicab trajectory data, and model the influence of latent business areas via ClusRanking. Also, we use a linear model to fuse these three influential factors and predict estate investment values. Moreover, we simultaneously consider individual, peer and zone dependencies, and derive an estate-specific ranking likelihood as the objective function. Finally, we conduct a comprehensive evaluation with real-world estate related data, and the experimental results demonstrate the effectiveness of our method.

Online social networks offering various services have become ubiquitous in our daily life. Meanwhile, users nowadays are usually involved in multiple online social networks simultaneously to enjoy specific services provided by different networks. Formally, social networks that share some common users are named as partially aligned networks. In this paper, we want to predict the formation of social links in multiple partially aligned social networks at the same time, which is formally defined as the multi-network link (formation) prediction problem. In multiple partially aligned social networks, users can be extensively correlated with each other by various connections. To categorize these diverse connections among users, 7 "intra-network social meta paths" and 4 categories of "inter-network social meta paths" are proposed in this paper. These "social meta paths" can cover a wide variety of connection information in the network, some of which can be helpful for solving the multi-network link prediction problem but some can be not. To utilize useful connection, a subset of the most informative "social meta paths" are picked, the process of which is formally defined as "social meta path selection" in this paper. An effective general link formation prediction framework, Mli (Multi-network Link Identifier), is proposed in this paper to solve the multi-network link (formation) prediction problem. Built with heterogenous topological features extracted based on the selected "social meta paths" in the multiple partially aligned social networks, Mli can help refine and disambiguate the prediction results reciprocally in all aligned networks. Extensive experiments conducted on real-world partially aligned heterogeneous networks, Foursquare and Twitter, demonstrate that Mli can solve the multi-network link prediction problem very well.

Support vector machine (SVM) has been one of the most popular learning algorithms, with the central idea of maximizing the minimum margin, i.e., the smallest distance from the instances to the classification boundary. Recent theoretical results, however, disclosed that maximizing the minimum margin does not necessarily lead to better generalization performances, and instead, the margin distribution has been proven to be more crucial. In this paper, we propose the Large margin Distribution Machine (LDM), which tries to achieve a better generalization performance by optimizing the margin distribution. We characterize the margin distribution by the first- and second-order statistics, i.e., the margin mean and variance. The LDM is a general learning approach which can be used in any place where SVM can be applied, and its superiority is verified both theoretically and empirically in this paper.

Multi-label learning studies the problem where each example is represented by a single instance while associated with a set of labels simultaneously. During the past decade, significant amount of progresses have been made toward this emerging machine learning paradigm. This paper aims to provide a timely review on this area with emphasis on state-of-the-art multi-label learning algorithms. Firstly, fundamentals on multi-label learning including formal definition and evaluation metrics are given. Secondly and primarily, eight representative multi-label learning algorithms are scrutinized under common notations with relevant analyses and discussions. Thirdly, several related learning settings are briefly summarized. As a conclusion, online resources and open research problems on multi-label learning are outlined for reference purposes.

In multi-label learning, it is rather expensive to label instances since they are simultaneously associated with multiple labels. Therefore, active learning, which reduces the labeling cost by actively querying the labels of the most valuable data, becomes particularly important for multi-label learning. A strong multi-label active learning algorithm usually consists of two crucial elements: a reasonable criterion to evaluate the gain of queried label, and an effective classification model, based on whose prediction the criterion can be accurately computed. In this paper, we first introduce an effective multi-label classification model by combining label ranking with threshold learning, which is incrementally trained to avoid retraining from scratch after every query. Based on this model, we then propose to exploit both uncertainty and diversity in the instance space as well as the label space, and actively query the instance-label pairs which can improve the classification model most. Experimental results demonstrate the superiority of the proposed approach to state-of-the-art methods.

Class-imbalance is very common in real data mining tasks. Previous studies focused on binary-class imbalance problem, whereas multi-class imbalance problem is more challenging. Error correcting output codes (ECOC) technique can be applied to class-imbalance problem, however, the standard ECOC aims at maximizing accuracy, ignoring the fact that, when class-imbalance is really a problem, the minority classes are more important than the majority classes. To enable ECOC to tackle multi-class imbalance, it is desired to have an appropriate code matrix, an effective learning strategy and a decoding strategy emphasizing the minority classes. In this paper, based on the aforementioned consideration, we propose the imECOC method which works on dichotomies to handle both the between-class imbalance and within-class imbalance. As the dichotomy classifiers contribute differently to the final prediction, imECOC assigns weights to dichotomies and uses weighted distance for decoding, where the optimal dichotomy weights are obtained by minimizing a weighted loss in favor of the minority classes. Experimental results on fourteen data sets show that, imECOC performs significantly better than many state-of-the-art multi-class imbalance learning methods, no matter whether multi-class F1, G-mean or AUC are used as evaluation measures.

Ever-increasing design complexity and advances of technology impose great challenges on the design of modern microprocessors. One such challenge is to determine promising microprocessor configurations to meet specific design constraints, which is called Design Space Exploration (DSE). In the computer architecture community, supervised learning techniques have been applied to DSE to build regression models for predicting the qualities of design configurations. For supervised learning, however, considerable simulation costs are required for attaining the labeled design configurations. Given limited resources, it is difficult to achieve high accuracy. In this article, inspired by recent advances in semisupervised learning and active learning, we propose the COAL approach which can exploit unlabeled design configurations to significantly improve the models. Empirical study demonstrates that COAL significantly outperforms a state-of-the-art DSE technique by reducing mean squared error by 35% to 95%, and thus, promising architectures can be attained more efficiently.

The problem of multilabel classification has attracted great interest in the last decade, where each instance can be assigned with a set of multiple class labels simultaneously. It has a wide variety of real-world applications, e.g., automatic image annotations and gene function analysis. Current research on multilabel classification focuses on supervised settings which assume existence of large amounts of labeled training data. However, in many applications, the labeling of multilabeled data is extremely expensive and time consuming, while there are often abundant unlabeled data available. In this paper, we study the problem of transductive multilabel learning and propose a novel solution, called Trasductive Multilabel Classification (TraM), to effectively assign a set of multiple labels to each instance. Different from supervised multilabel learning methods, we estimate the label sets of the unlabeled instances effectively by utilizing the information from both labeled and unlabeled data. We first formulate the transductive multilabel learning as an optimization problem of estimating label concept compositions. Then, we derive a closed-form solution to this optimization problem and propose an effective algorithm to assign label sets to the unlabeled instances. Empirical studies on several real-world multilabel learning tasks demonstrate that our TraM method can effectively boost the performance of multilabel classification by using both labeled and unlabeled data.

Multi-label learning arises in many real-world tasks where an object is naturally associated with multiple concepts. It is well-accepted that, in order to achieve a good performance, the relationship among labels should be exploited. Most existing approaches require the label relationship as prior knowledge, or exploit by counting the label co-occurrence. In this paper, we propose the MAHR approach, which is able to automatically discover and exploit label relationship. Our basic idea is that, if two labels are related, the hypothesis generated for one label can be helpful for the other label. MAHR implements the idea as a boosting approach with a hypothesis reuse mechanism. In each boosting round, the base learner for a label is generated by not only learning on its own task but also reusing the hypotheses from other labels, and the amount of reuse across labels provides an estimate of the label relationship. Extensive experimental results validate that MAHR is able to achieve superior performance and discover reasonable label relationship. Moreover, we disclose that the label relationship is usually asymmetric.

No abstract available.

Anomalies are data points that are few and different. As a result of these properties, we show that, anomalies are susceptible to a mechanism called isolation . This article proposes a method called Isolation Forest ( i Forest), which detects anomalies purely based on the concept of isolation without employing any distance or density measure---fundamentally different from all existing methods. As a result, i Forest is able to exploit subsampling (i) to achieve a low linear time-complexity and a small memory-requirement and (ii) to deal with the effects of swamping and masking effectively. Our empirical evaluation shows that i Forest outperforms ORCA, one-class SVM, LOF and Random Forests in terms of AUC, processing time, and it is robust against masking and swamping effects. i Forest also works well in high dimensional problems containing a large number of irrelevant attributes, and when anomalies are not available in training sample.

Advances in GPS tracking technology have enabled us to install GPS tracking devices in city taxis to collect a large amount of GPS traces under operational time constraints. These GPS traces provide unparallel opportunities for us to uncover taxi driving fraud activities. In this paper, we develop a taxi driving fraud detection system, which is able to systematically investigate taxi driving fraud. In this system, we first provide functions to find two aspects of evidences: travel route evidence and driving distance evidence. Furthermore, a third function is designed to combine the two aspects of evidences based on Dempster-Shafer theory. To implement the system, we first identify interesting sites from a large amount of taxi GPS logs. Then, we propose a parameter-free method to mine the travel route evidences. Also, we introduce route mark to represent a typical driving path from an interesting site to another one. Based on route mark, we exploit a generative statistical model to characterize the distribution of driving distance and identify the driving distance evidences. Finally, we evaluate the taxi driving fraud detection system with large scale real-world taxi GPS logs. In the experiments, we uncover some regularity of driving fraud activities and investigate the motivation of drivers to commit a driving fraud by analyzing the produced taxi fraud data.

Cotraining, a paradigm of semisupervised learning, is promised to alleviate effectively the shortage of labeled examples in supervised learning. The standard two-view cotraining requires the data set to be described by two views of features, and previous studies have shown that cotraining works well if the two views satisfy the sufficiency and independence assumptions. In practice, however, these two assumptions are often not known or ensured (even when the two views are given). More commonly, most supervised data sets are described by one set of attributes (one view). Thus, they need be split into two views in order to apply the standard two-view cotraining. In this paper, we first propose a novel approach to empirically verify the two assumptions of cotraining given two views. Then, we design several methods to split single view data sets into two views, in order to make cotraining work reliably well. Our empirical results show that, given a whole or a large labeled training set, our view verification and splitting methods are quite effective. Unfortunately, cotraining is called for precisely when the labeled training set is small. However, given small labeled training sets, we show that the two cotraining assumptions are difficult to verify, and view splitting is unreliable. Our conclusions for cotraining's effectiveness are mixed. If two views are given, and known to satisfy the two assumptions, cotraining works well. Otherwise, based on small labeled training sets, verifying the assumptions or splitting single view into two views are unreliable; thus, it is uncertain whether the standard cotraining would work or not.

The increasing availability of large-scale location traces creates unprecedent opportunities to change the paradigm for identifying abnormal moving activities. Indeed, various aspects of abnormality of moving patterns have recently been exploited, such as wrong direction and wandering. However, there is no recognized way of combining different aspects into an unified evolving abnormality score which has the ability to capture the evolving nature of abnormal moving trajectories. To that end, in this paper, we provide an evolving trajectory outlier detection method, named TOP-EYE, which continuously computes the outlying score for each trajectory in an accumulating way. Specifically, in TOP-EYE, we introduce a decay function to mitigate the influence of the past trajectories on the evolving outlying score, which is defined based on the evolving moving direction and density of trajectories. This decay function enables the evolving computation of accumulated outlying scores along the trajectories. An advantage of TOP-EYE is to identify evolving outliers at very early stage with relatively low false alarm rate. Finally, experimental results on real-world location traces show that TOP-EYE can effectively capture evolving abnormal trajectories.

Ensemble learning aims to improve generalization ability by using multiple base learners. It is well-known that to construct a good ensemble, the base learners should be accurate as well as diverse. In this paper, unlabeled data is exploited to facilitate ensemble learning by helping augment the diversity among the base learners. Specifically, a semi-supervised ensemble method named UDEED is proposed. Unlike existing semi-supervised ensemble methods where error-prone pseudo-labels are estimated for unlabeled data to enlarge the labeled data to improve accuracy, UDEED works by maximizing accuracies of base learners on labeled data while maximizing diversity among them on unlabeled data. Experiments show that UDEED can effectively utilize unlabeled data for ensemble learning and is highly competitive to well-established semi-supervised ensemble methods.

Existing cost-sensitive learning methods require that the unequal misclassification costs should be given as precise values. In many real-world applications, however, it is generally difficult to have a precise cost value since the user maybe only knows that one type of mistake is much more severe than another type, yet it is infeasible to give a precise description. In such situations, it is more meaningful to work with a cost interval instead of a precise cost value. In this paper we report the first study along this direction. We propose the CISVM method, a support vector machine, to work with cost interval information. Experiments show that when there are only cost intervals available, CISVM is significantly superior to standard cost-sensitive SVMs using any of the minimal cost, mean cost and maximal cost to learn. Moreover, considering that in some cases other information about costs can be obtained in addition to cost intervals, such as the distribution of costs, we propose a general approach CODIS for using the distribution information to help improve performance. Experiments show that this approach can reduce 60% more risks than the standard cost-sensitive SVM which assumes the expected cost is the true value.

Multilabel learning deals with data associated with multiple labels simultaneously. Like other data mining and machine learning tasks, multilabel learning also suffers from thecurse of dimensionality. Dimensionality reduction has been studied for many years, however, multilabel dimensionality reduction remains almost untouched. In this article, we propose a multilabel dimensionality reduction method, MDDM, with two kinds of projection strategies, attempting to project the original data into a lower-dimensional feature space maximizing the dependence between the original feature description and the associated class labels. Based on the Hilbert-Schmidt Independence Criterion, we derive a eigen-decomposition problem which enables the dimensionality reduction process to be efficient. Experiments validate the performance of MDDM.

It is well known that the key of Bayesian classifier learning is to balance the two important issues, that is, the exploration of attribute dependencies in high orders for ensuring a sufficient flexibility in approximating the ground-truth dependencies, and the exploration of low orders for ensuring a stable probability estimate from limited training samples. By allowing one-order attribute dependencies, one-dependence estimators (ODEs) have been shown to be able to approximate the ground-truth attribute dependencies whilst keeping the effectiveness of probability estimation, and therefore leading to excellent performance. In previous studies, however, ODEs were exploited in simple ways, such as by averaging, for classification. In this paper, we propose a semi-naive exploitation of ODEs that fits a function of ODEs to pursue higher-order attribute dependencies. Extensive experiments show that the proposed SNODE approach can achieve better performance than many state-of-the-art Bayesian classifiers.

Linear discriminant analysis (LDA) is a well-known dimension reduction approach, which projects high-dimensional data into a low-dimensional space with the best separation of different classes. In many tasks, the data accumulates over time, and thus incremental LDA is more desirable than batch LDA. Several incremental LDA algorithms have been developed and achieved success; however, the eigen-problem involved requires a large computation cost, which hampers the efficiency of these algorithms. In this paper, we propose a new incremental LDA algorithm, LS-ILDA, based on the least square solution of LDA. When new samples are received, LS-ILDA incrementally updates the least square solution of LDA. Our analysis discloses that this algorithm produces the exact least square solution of batch LDA, while its computational cost is O(min(n, d) × d) for one update on dataset containing n instances in d-dimensional space. Experimental results show that comparing with state-of-the-art incremental LDA algorithms, our proposed LS-ILDA achieves high accuracy with low time cost.

The Drosophila gene expression pattern images document the spatial and temporal dynamics of gene expression and they are valuable tools for explicating the gene functions, interaction, and networks during Drosophila embryogenesis. To provide text-based pattern searching, the images in the Berkeley Drosophila Genome Project (BDGP) study are annotated with ontology terms manually by human curators. We present a systematic approach for automating this task, because the number of images needing text descriptions is now rapidly increasing. We consider both improved feature representation and novel learning formulation to boost the annotation performance. For feature representation, we adapt the bag-of-words scheme commonly used in visual recognition problems so that the image group information in the BDGP study is retained. Moreover, images from multiple views can be integrated naturally in this representation. To reduce the quantization error caused by the bag-of-words representation, we propose an improved feature representation scheme based on the sparse learning technique. In the design of learning formulation, we propose a local regularization framework that can incorporate the correlations among terms explicitly. We further show that the resulting optimization problem admits an analytical solution. Experimental results show that the representation based on sparse learning outperforms the bag-of-words representation significantly. Results also show that incorporation of the term-term correlations improves the annotation performance consistently.

Text categorization is the task of assigning predefined categories to natural language text. With the widely used 'bag of words' representation, previous researches usually assign a word with values such that whether this word appears in the document concerned or how frequently this word appears. Although these values are useful for text categorization, they have not fully expressed the abundant information contained in the document. This paper explores the effect of other types of values, which express the distribution of a word in the document. These novel values assigned to a word are called distributional features, which include the compactness of the appearances of the word and the position of the first appearance of the word. The proposed distributional features are exploited by a tf idf style equation and different features are combined using ensemble learning techniques. Experiments show that the distributional features are useful for text categorization. In contrast to using the traditional term frequency values solely, including the distributional features requires only a little additional cost, while the categorization performance can be significantly improved. Further analysis shows that the distributional features are especially useful when documents are long and the writing style is casual.

Most existing model-based approaches to anomaly detection construct a profile of normal instances, then identify instances that do not conform to the normal profile as anomalies. This paper proposes a fundamentally different model-based method that explicitly isolates anomalies instead of profiles normal points. To our best knowledge, the concept of isolation has not been explored in current literature. The use of isolation enables the proposed method, iForest, to exploit sub-sampling to an extent that is not feasible in existing methods, creating an algorithm which has a linear time complexity with a low constant and a low memory requirement. Our empirical evaluation shows that iForest performs favourably to ORCA, a near-linear time complexity distance-based method, LOF and random forests in terms of AUC and processing time, and especially in large data sets. iForest also works well in high dimensional problems which have a large number of irrelevant attributes, and in situations where training set does not contain any anomalies.

With the rapid evolution of Internet applications, people all over the world are sharing pictures, videos and audios online, and thus, content-based analysis is often demanded. Test efficiency is crucial to the success of online information processing. One obstacle to high-speed testing is the time cost of feature extraction for test objects, particularly for objects with complex representation such as images, videos and audios. In this paper, we study the problem of reducing test time cost by extracting cheap but sufficient features. We propose the TEFE (time-efficient feature extraction) approach, which balances between the test accuracy and test time cost by extracting a proper subset of features for each test object. In the implementation, TEFE trains a sequence of support vector machines and classifies each test object cascadingly. Empirical study shows that TEFE is time efficient while holding a classification accuracy close to that of using all features. It also shows that the test time is linearly adjustable in TEFE.

Multi-instance multi-label learning (MIML) deals with the problem where each training example is associated with not only multiple instances but also multiple class labels. Previous MIML algorithms work by identifying its equivalence in degenerated versions of multi-instance multi-label learning. However, useful information encoded in training examples may get lost during the identification process. In this paper, a maximum margin method is proposed for MIML which directly exploits the connections between instances and labels. The learning task is formulated as a quadratic programming (QP) problem and implemented in its dual form. Applications to scene classification and text categorization show that the proposed approach achieves superior performance over existing MIML methods.

This paper is motivated to improve the performance of individual ensembles using a hybrid mechanism in the regression setting. Based on an error-ambiguity decomposition, we formally analyze the optimal linear combination of two base ensembles, which is then extended to multiple individual ensembles via pairwise combinations. The Cocktail ensemble approach is proposed based on this analysis. Experiments over a broad range of data sets show that the proposed approach outperforms the individual ensembles, two other methods of ensemble combination, and two state-of-the-art regression approaches.

Dimensionality reduction is among the keys in mining high-dimensional data. This paper studies semi-supervised dimensionality reduction. In this setting, besides abundant unlabeled examples, domain knowledge in the form of pairwise constraints are available, which specifies whether a pair of instances belong to the same class (must-link constraints) or different classes (cannot-link constraints). We propose the SSDR algorithm, which can preserve the intrinsic structure of the unlabeled data as well as both the must-link and cannot-link constraints defined on the labeled examples in the projected low-dimensional space. The SSDR algorithm is efficient and has a closed form solution. Experiments on a broad range of data sets show that SSDR is superior to many established dimensionality reduction methods.

The traditional setting of supervised learning requires a large amount of labeled training examples in order to achieve good generalization. However, in many practical applications, unlabeled training examples are readily available, but labeled ones are fairly expensive to obtain. Therefore, semisupervised learning has attracted much attention. Previous research on semisupervised learning mainly focuses on semisupervised classification. Although regression is almost as important as classification, semisupervised regression is largely understudied. In particular, although cotraining is a main paradigm in semisupervised learning, few works has been devoted to cotraining-style semisupervised regression algorithms. In this paper, a cotraining-style semisupervised regression algorithm, that is, COREG, is proposed. This algorithm uses two regressors, each labels the unlabeled data for the other regressor, where the confidence in labeling an unlabeled example is estimated through the amount of reduction in mean squared error over the labeled neighborhood of that example. Analysis and experiments show that COREG can effectively exploit unlabeled data to improve regression estimates.

Under-sampling is a class-imbalance learning method which uses only a subset of major class examples and thus is very efficient. The main deficiency is that many major class examples are ignored. We propose two algorithms to overcome the deficiency. EasyEnsemble samples several subsets from the major class, trains a learner using each of them, and combines the outputs of those learners. BalanceCascade is similar to EasyEnsemble except that it removes correctly classified major class examples of trained learners from further consideration. Experiments show that both of the proposed algorithms have better AUC scores than many existing class-imbalance learning methods. Moreover, they have approximately the same training time as that of under-sampling, which trains significantly faster than other methods.

In real-world applications the number of examples in one class may overwhelm the other class, but the primary interest is usually on the minor class. Cost-sensitive learning has been deeded as a good solution to these class-imbalanced tasks, yet it is not clear how does the class-imbalance affect cost-sensitive classifiers. This paper presents an empirical study using 38 data sets, which discloses that class-imbalance often affects the performance of cost-sensitive classifiers: When the misclassification costs are not seriously unequal, cost-sensitive classifiers generally favor natural class distribution although it might be imbalanced; while when misclassification costs are seriously unequal, a balanced class distribution is more favorable.

Choosing an appropriate kernel is one of the key problems in kernel-based methods. Most existing kernel selection methods require that the class labels of the training examples are known. In this paper, we propose an adaptive kernel selection method for kernel principal component analysis, which can effectively learn the kernels when the class labels of the training examples are not available. By iteratively optimizing a novel criterion, the proposed method can achieve nonlinear feature extraction and unsupervised kernel learning simultaneously. Moreover, a non-iterative approximate algorithm is developed. The effectiveness of the proposed algorithms are validated on UCI datasets and the COIL-20 object recognition database.

Similarity measure is one of the keys of a high- performance content-based image retrieval (CBIR) system. Given a pair of images, existing similarity measures usually produce a static and constant similarity score. However, an image can usually be perceived with different meanings and therefore, the similarity between the same pair of images may change when the concept being queried changes. This paper proposes a query-sensitive similarity measure, Qsim, which takes the concept being queried into account in measuring image similarities, by exploiting the query image as well as the images labeled by user in the relevance feedback process. Experimental comparisons to state-of-the-art techniques show that Qsim has superior performance.

In multilabel learning, each instance in the training set is associated with a set of labels and the task is to output a label set whose size is unknown a priori for each unseen instance. In this paper, this problem is addressed in the way that a neural network algorithm named BP-MLL, i.e., Backpropagation for Multilabel Learning, is proposed. It is derived from the popular Backpropogation algorithm through employing a novel error function capturing the characteristics of multilabel learning, i.e., the labels belonging to an instance should be ranked higher than those not belonging to that instance. Applications to two real-world multilabel learning problems, i.e., functional genomics and text categorization, show that the performance of BP-MLL is superior to that of some well-established multilabel learning algorithms.

This paper studies empirically the effect of sampling and threshold-moving in training cost-sensitive neural networks. Both oversampling and undersampling are considered. These techniques modify the distribution of the training data such that the costs of the examples are conveyed explicitly by the appearances of the examples. Threshold-moving tries to move the output threshold toward inexpensive classes such that examples with higher costs become harder to be misclassified. Moreover, hard-ensemble and soft-ensemble, i.e., the combination of above techniques via hard or soft voting schemes, are also tested. Twenty-one UCl data sets with three types of cost matrices and a real-world cost-sensitive data set are used in the empirical study. The results suggest that cost-sensitive learning with multiclass tasks is more difficult than with two-class tasks, and a higher degree of class imbalance may increase the difficulty. It also reveals that almost all the techniques are effective on two-class tasks, while most are ineffective and even may cause negative effect on multiclass tasks. Overall, threshold-moving and soft-ensemble are relatively good choices in training cost-sensitive neural networks. The empirical study also suggests that some methods that have been believed to be effective in addressing the class imbalance problem may, in fact, only be effective on learning with imbalanced two-class data sets.

Relevance feedback is an effective scheme bridging the gap between high-level semantics and low-level features in content-based image retrieval (CBIR). In contrast to previous methods which rely on labeled images provided by the user, this article attempts to enhance the performance of relevance feedback by exploiting unlabeled images existing in the database. Concretely, this article integrates the merits of semisupervised learning and active learning into the relevance feedback process. In detail, in each round of relevance feedback two simple learners are trained from the labeled data, that is, images from user query and user feedback. Each learner then labels some unlabeled images in the database for the other learner. After retraining with the additional labeled data, the learners reclassify the images in the database and then their classifications are merged. Images judged to be positive with high confidence are returned as the retrieval result, while those judged with low confidence are put into thepoolwhich is used in the next round of relevance feedback. Experiments show that using semisupervised learning and active learning simultaneously in CBIR is beneficial, and the proposed method achieves better performance than some existing methods.

In many practical data mining applications, such as Web page classification, unlabeled training examples are readily available, but labeled ones are fairly expensive to obtain. Therefore, semi-supervised learning algorithms such as co-training have attracted much attention. In this paper, a new co-training style semi-supervised learning algorithm, named tri-training, is proposed. This algorithm generates three classifiers from the original labeled example set. These classifiers are then refined using unlabeled examples in the tri-training process. In detail, in each round of tri-training, an unlabeled example is labeled for a classifier if the other two classifiers agree on the labeling, under certain conditions. Since tri-training neither requires the instance space to be described with sufficient and redundant views nor does it put any constraints on the supervised learning algorithm, its applicability is broader than that of previous co-training style algorithms. Experiments on UCI data sets and application to the Web page classification task indicate that tri-training can effectively exploit unlabeled data to enhance the learning performance.

Decision tree is with good comprehensibility while neural network ensemble is with strong generalization ability. These merits are integrated into a novel decision tree algorithm NeC4.5. This algorithm trains a neural network ensemble at first. Then, the trained ensemble is employed to generate a new training set through replacing the desired class labels of the original training examples with those output from the trained ensemble. Some extra training examples are also generated from the trained ensemble and added to the new training set. Finally, a C4.5 decision tree is grown from the new training set. Since its learning results are decision trees, the comprehensibility of NeC4.5 is better than that of neural network ensemble. Moreover, experiments show that the generalization ability of NeC4.5 decision trees can be better than that of C4.5 decision trees.