Online Classification Tasks Research Articles

Online Passive-Aggressive Active Learning for Trapezoidal Data Streams.

The idea of combining the active query strategy and the passive-aggressive (PA) update strategy in online learning can be credited to the PA active (PAA) algorithm, which has proven to be effective in learning linear classifiers from datasets with a fixed feature space. We propose a novel family of online active learning algorithms, named PAA learning for trapezoidal data streams (PAA [Formula: see text] ) and multiclass PAA [Formula: see text] (MPAA [Formula: see text] ) (and their variants), for binary and multiclass online classification tasks on trapezoidal data streams where the feature space may expand over time. Under the context of an ever-changing feature space, we provide the theoretical analysis of the mistake bounds for both PAA [Formula: see text] and MPAA [Formula: see text] . Our experiments on a wide variety of benchmark datasets have confirm that the combination of the instance-regulated active query strategy and the PA update strategy is much more effective in learning from trapezoidal data streams. We have also compared PAA [Formula: see text] with online learning with streaming features (OL [Formula: see text] )-the state-of-the-art approach in learning linear classifiers from trapezoidal data streams. PAA [Formula: see text] could achieve much better classification accuracy, especially for large-scale real-world data streams.

SWINN: Efficient nearest neighbor search in sliding windows using graphs

Nearest neighbor search (NNS) is one of the main concerns in data stream applications since similarity queries can be used in multiple scenarios. Online NNS is usually performed on a sliding window by lazily scanning every element currently stored in the window. This paper proposes Sliding Window-based Incremental Nearest Neighbors (SWINN), a graph-based online search index algorithm for speeding up NNS in potentially never-ending and dynamic data stream tasks. Our proposal broadens the application of online NNS-based solutions, as even moderately large data buffers become impractical to handle when a naive NNS strategy is selected. SWINN enables efficient handling of large data buffers by using an incremental strategy to build and update a search graph supporting any distance metric. Vertices can be added and removed from the search graph. To keep the graph reliable for search queries, lightweight graph maintenance routines are run. According to experimental results, SWINN is significantly faster than performing a naive complete scan of the data buffer while keeping competitive search recall values. We also apply SWINN to online classification and regression tasks and show that our proposal is effective against popular online machine learning algorithms.

Consensus Building in On-Line Citizen Science

A number of initiatives invite members of the public to perform online classification tasks such as identifying objects in images. These tasks are crucial to numerous large-scale Citizen Science projects in different disciplines, with volunteers using their knowledge and online support tools to, for example, identify species of wildlife or classify galaxies by their shapes. However, for complex classification tasks, such as this case study on identifying species of bumblebee, reaching an agreement between volunteers - or even between experts~-~may require consensus-building processes. Collaboration and teamwork approaches to problem solving and decision-making have been widely documented to improve both task performance and user learning in the real world. Most of these processes and projects are mediated online through feedback delivered in an asynchronous manner, and this article thus addresses a central research question: How do participants involved in species identification tasks respond to different forms of feedback provided in online collaboration, designed to support peer-learning and improve task performance? We tested four different approaches to feedback within a collaboration task, where participants reviewed their previously annotated data based on information curated from their peers on a long running online citizen science initiative. The selected interfaces have a strong foundation in social science and psychology literature and can be applied to citizen science practices as well as other online communities. Results showed that while all four approaches increased accuracy, there were differences based on the types of consensus that existed before collaboration. Such differences highlight the usefulness of different forms of feedback during collaboration for increasing data accuracy of identification and furthering users' expertise on identification tasks. We found that anonymised and goal-directed free text comments posted on social learning interfaces were most effective in improving data accuracy as well as creating opportunities for peer-learning, particularly where the species identification task was more difficult. This study has significant implications for extending the practice of citizen science across formal and informal learning environments and reaching out to a variety of users.

The development of on-line surface defect detection system for jujubes based on hyperspectral images

• A hyperspectral system was built for collecting hyperspectral image data and performing skin defect detection of jujubes online. • Hyperspectral data of jujubes with common skin defects were obtained for training classification models. • Support vector machine and artificial neural network models were used for online skin defect detection of jujubes. • To reduce the computation time of online classification, spectral bands were selected with equal band interval method or the principal component analysis. • Experimental results showed that artificial neural network models using 14 bands can compromise the accuracy and speed for online classification of jujubes. This paper presents the development of an on-line surface defect system using hyperspectral images for jujubes. A push-broom hyperspectral system was built for collecting hyperspectral image data and detecting skin defects of jujube online. Hyperspectral images with an effective wavelength range of 468–950 nm were obtained for jujubes with normal surface or common skin defect types (i.e., russeting, decay, white fungus, black fungus and crack). Support vector machine (SVM) and artificial neural networks (ANN) models were used to classify surface defects of jujubes. The classification accuracies, with the use of full wavelength range, of ANN and SVM models for jujube skin defects are 96.5% and 96.3% respectively. The times required for processing one jujube face are about 25 and 320 s for ANN and SVM models respectively. To reduce the computation time of online classification tasks, spectral bands were selected from a wavelength range of 468–760 nm with equal band intervals or by the principal component analysis (PCA) method. Experimental results showed that the accuracy of SVM and ANN models using 14 bands (469, 491, 513, 535, 558, 580, 602, 624, 646, 668, 691, 713, 735 and 757 nm), selected by equal wavelength intervals, were 94.4% and 95% respectively. And the accuracies of ANN and SVM models with 14 bands (470, 493, 534, 555, 590, 623, 632, 654, 672, 674, 683, 696, 707 and 747) selected by PCA are 95% and 94.6% respectively. The classification time, with the use of 14 bands, of ANN and SVM models for jujube skin defects reduced to 16.6 and 30 s respectively. The online line scanning and classification hyperspectral imaging system can be used for surface defect detection of other fruits.

Second-Order Online Active Learning and Its Applications

The goal of online active learning is to learn predictive models from a sequence of unlabeled data given limited label query budget. Unlike conventional online learning tasks, online active learning is considerably more challenging because of two reasons. First, it is difficult to design an effective query strategy to decide when is appropriate to query the label of an incoming instance given limited query budget. Second, it is also challenging to decide how to update the predictive models effectively whenever the true label of an instance is queried. Most existing approaches for online active learning are often based on a family of first-order online learning algorithms, which are simple and efficient but fall short in the slow convergence and sub-optimal solution in exploiting the labeled training data. To solve these issues, this paper presents a novel framework of Second-order Online Active Learning (SOAL) by fully exploiting both the first-order and second-order information. The proposed algorithms are able to achieve effective online learning efficacy, maximize the predictive accuracy, and minimize the labeling cost. To make SOAL more practical for real-world applications, especially for class-imbalanced online classification tasks (e.g., malicious web detection), we extend the SOAL framework by proposing the Cost-sensitive Second-order Online Active Learning algorithm named “SOAL $_{CS}$ ”, which is devised by maximizing the sum of weighted sensitivity and specificity or minimizing the cost of weighted mistakes of different classes. We conducted both theoretical analysis and empirical studies, including an extensive set of experiments on a variety of large-scale real-world datasets, in which the promising empirical results validate the efficacy and scalability of the proposed algorithms towards large-scale online learning tasks.

Large-Scale Kernel-Based Feature Extraction via Low-Rank Subspace Tracking on a Budget

Kernel-based methods enjoy powerful generalization capabilities in learning a variety of pattern recognition tasks. When such methods are provided with sufficient training data, broadly applicable classes of nonlinear functions can be approximated with desired accuracy. Nevertheless, inherent to the nonparametric nature of kernel-based estimators are computational and memory requirements that become prohibitive with large-scale datasets. In response to this formidable challenge, this paper puts forward a low-rank, kernel-based, feature extraction approach that is particularly tailored for online operation. A novel generative model is introduced to approximate high-dimensional (possibly infinite) features via a low-rank nonlinear subspace, the learning of which lends itself to a kernel function approximation. Offline and online solvers are developed for the subspace learning task, along with affordable versions, in which the number of stored data vectors is confined to a predefined budget. Analytical results provide performance bounds on how well the kernel matrix as well as kernel-based classification and regression tasks can be approximated by leveraging budgeted online subspace learning and feature extraction schemes. Tests on synthetic and real datasets demonstrate and benchmark the efficiency of the proposed method for dynamic nonlinear subspace tracking as well as online classification and regressions tasks.

Automatic emotion detection in text streams by analyzing Twitter data

Techniques to detect the emotions expressed in microblogs and social media posts have a wide range of applications including, detecting psychological disorders such as anxiety or depression in individuals or measuring the public mood of a community. A major challenge for automated emotion detection is that emotions are subjective concepts with fuzzy boundaries and with variations in expression and perception. To address this issue, a dimensional model of affect is utilized to define emotion classes. Further, a soft classification approach is proposed to measure the probability of assigning a message to each emotion class. We develop and evaluate a supervised learning system to automatically classify emotion in text stream messages. Our approach includes two main tasks: an offline training task and an online classification task. The first task creates models to classify emotion in text messages. For the second task, we develop a two-stage framework called EmotexStream to classify live streams of text messages for the real-time emotion tracking. Moreover, we propose an online method to measure public emotion and detect emotion burst moments in live text streams.

Online Passive-Aggressive Active learning

We investigate online active learning techniques for online classification tasks. Unlike traditional supervised learning approaches, either batch or online learning, which often require to request class labels of each incoming instance, online active learning queries only a subset of informative incoming instances to update the classification model, aiming to maximize classification performance with minimal human labelling effort during the entire online learning task. In this paper, we present a new family of online active learning algorithms called Passive-Aggressive Active (PAA) learning algorithms by adapting the Passive-Aggressive algorithms in online active learning settings. Unlike conventional Perceptron-based approaches that employ only the misclassified instances for updating the model, the proposed PAA learning algorithms not only use the misclassified instances to update the classifier, but also exploit correctly classified examples with low prediction confidence. Specifically, we propose several variants of PAA algorithms to tackle three types of online learning tasks: binary classification, multi-class classification, and cost-sensitive classification. We give the mistake bounds of the proposed algorithms in theory, and conduct extensive experiments to evaluate the empirical performance of our techniques on both standard and large-scale datasets, in which the encouraging results validate the empirical effectiveness of the proposed algorithms.

Two approaches for novelty detection using random forest

In many online classification tasks or non-exhaustive learning, it is often impossible to define a training set with a complete set of classes. The presence of new classes as well as the novelties caused by data errors can severely affect the performance of classifiers. Traditional proximity-based approaches usually utilize the distance to measure the proximity of different samples. In this study, we propose a framework that uses ensemble learning to detect novelty based on Random Forest (RF). The proposed framework is based on the observation that an ensemble of classifiers can provide a kind of metric to characterize different classes and measure their proximity. In particular, we apply ensemble methods with the decision tree as base classifiers and present two specific approaches, RFV and RFP, based on random forest. RFV uses the vote distribution of RF on a testing sample, and RFP takes the proximity matrix of RF as a special kernel metric to discover the novelty. The proposed approaches are compared against two common approaches: support vector domain description (SVDD) and Gaussian Mixed Model (GMM) on one artificial data set and five benchmark data sets. The experimental results show that the proposed methods achieve better performance in terms of accuracy and recall.

Development of a Mechatronics-Based Citizen Science Platform for Aquatic Environmental Monitoring

In this paper, we present the design and proof of concept of a low-cost, self-sustained mobile surface vehicle for environmental monitoring. The vehicle is equipped with water quality sensors and cameras for image acquisition and two thrusters allowing it to maneuver. The device is part of a citizen science system wherein volunteers help in the environmental recovery of polluted bodies of water by performing online research and classification tasks. While aiding in environmental recovery, the users become more proficient in wildlife classification and water quality assessment. Telemetry collected by the vehicle is wirelessly uploaded to a web-based interface allowing volunteers to access images, positioning information, and water quality sensor data. The citizen science system allows for the study of social networking and human–computer interaction based on user activity. Specifically, the platform can be utilized to experiment with different interfaces and social interactions in hypothesis-driven research on human–computer interactions.

Cost-Sensitive Online Classification

Both cost-sensitive classification and online learning have been extensively studied in data mining and machine learning communities, respectively. However, very limited study addresses an important intersecting problem, that is, “Cost-Sensitive Online Classification". In this paper, we formally study this problem, and propose a new framework for Cost-Sensitive Online Classification by directly optimizing cost-sensitive measures using online gradient descent techniques. Specifically, we propose two novel cost-sensitive online classification algorithms, which are designed to directly optimize two well-known cost-sensitive measures: (i) maximization of weighted sum of sensitivity and specificity, and (ii) minimization of weighted misclassification cost. We analyze the theoretical bounds of the cost-sensitive measures made by the proposed algorithms, and extensively examine their empirical performance on a variety of cost-sensitive online classification tasks. Finally, we demonstrate the application of the proposed technique for solving several online anomaly detection tasks, showing that the proposed technique could be a highly efficient and effective tool to tackle cost-sensitive online classification tasks in various application domains.

Online Transfer Learning

In this paper, we propose a novel machine learning framework called “Online Transfer Learning” (OTL), which aims to attack an online learning task on a target domain by transferring knowledge from some source domain. We do not assume data in the target domain follows the same distribution as that in the source domain, and the motivation of our work is to enhance a supervised online learning task on a target domain by exploiting the existing knowledge that had been learnt from training data in source domains. OTL is in general very challenging since data in both source and target domains not only can be different in their class distributions, but also can be diverse in their feature representations. As a first attempt to this new research problem, we investigate two different settings of OTL: (i) OTL on homogeneous domains of common feature space, and (ii) OTL across heterogeneous domains of different feature spaces. For each setting, we propose effective OTL algorithms to solve online classification tasks, and show some theoretical bounds of the algorithms. In addition, we also apply the OTL technique to attack the challenging online learning tasks with concept-drifting data streams. Finally, we conduct extensive empirical studies on a comprehensive testbed, in which encouraging results validate the efficacy of our techniques.

An Incremental Classification Algorithm for Mining Data with Feature Space Heterogeneity

Feature space heterogeneity often exists in many real world data sets so that some features are of different importance for classification over different subsets. Moreover, the pattern of feature space heterogeneity might dynamically change over time as more and more data are accumulated. In this paper, we develop an incremental classification algorithm, Supervised Clustering for Classification with Feature Space Heterogeneity (SCCFSH), to address this problem. In our approach, supervised clustering is implemented to obtain a number of clusters such that samples in each cluster are from the same class. After the removal of outliers, relevance of features in each cluster is calculated based on their variations in this cluster. The feature relevance is incorporated into distance calculation for classification. The main advantage of SCCFSH lies in the fact that it is capable of solving a classification problem with feature space heterogeneity in an incremental way, which is favorable for online classification tasks with continuously changing data. Experimental results on a series of data sets and application to a database marketing problem show the efficiency and effectiveness of the proposed approach.

Incremental accelerated gradient methods for SVM classification: study of the constrained approach

We investigate constrained first order techniques for training support vector machines (SVM) for online classification tasks. The methods exploit the structure of the SVM training problem and combine ideas of incremental gradient technique, gradient acceleration and successive simple calculations of Lagrange multipliers. Both primal and dual formulations are studied and compared. Experiments show that the constrained incremental algorithms working in the dual space achieve the best trade-off between prediction accuracy and training time. We perform comparisons with an unconstrained large scale learning algorithm (Pegasos stochastic gradient) to emphasize that our choice can remain competitive for large scale learning due to the very special structure of the training problem.

Exploring the Trade-off Between Accuracy and Observational Latency in Action Recognition

An important aspect in designing interactive, action-based interfaces is reliably recognizing actions with minimal latency. High latency causes the system's feedback to lag behind user actions and thus significantly degrades the interactivity of the user experience. This paper presents algorithms for reducing latency when recognizing actions. We use a latency-aware learning formulation to train a logistic regression-based classifier that automatically determines distinctive canonical poses from data and uses these to robustly recognize actions in the presence of ambiguous poses. We introduce a novel (publicly released) dataset for the purpose of our experiments. Comparisons of our method against both a Bag of Words and a Conditional Random Field (CRF) classifier show improved recognition performance for both pre-segmented and online classification tasks. Additionally, we employ GentleBoost to reduce our feature set and further improve our results. We then present experiments that explore the accuracy/latency trade-off over a varying number of actions. Finally, we evaluate our algorithm on two existing datasets.

Sliding Window Generalized Kernel Affine Projection Algorithm Using Projection Mappings

Very recently, a solution to the kernel-based online classification problem has been given by the adaptive projected subgradient method (APSM). The developed algorithm can be considered as a generalization of a kernel affine projection algorithm (APA) and the kernel normalized least mean squares (NLMS). Furthermore, sparsification of the resulting kernel series expansion was achieved by imposing a closed ball (convex set) constraint on the norm of the classifiers. This paper presents another sparsification method for the APSM approach to the online classification task by generating a sequence of linear subspaces in a reproducing kernel Hilbert space (RKHS). To cope with the inherent memory limitations of online systems and to embed tracking capabilities to the design, an upper bound on the dimension of the linear subspaces is imposed. The underlying principle of the design is the notion of projection mappings. Classification is performed by metric projection mappings, sparsification is achieved by orthogonal projections, while the online system's memory requirements and tracking are attained by oblique projections. The resulting sparsification scheme shows strong similarities with the classical sliding window adaptive schemes. The proposed design is validated by the adaptive equalization problem of a nonlinear communication channel, and is compared with classical and recent stochastic gradient descent techniques, as well as with the APSM's solution where sparsification is performed by a closed ball constraint on the norm of the classifiers.

The Forgetron: A Kernel-Based Perceptron on a Budget

The Perceptron algorithm, despite its simplicity, often performs well in online classification tasks. The Perceptron becomes especially effective when it is used in conjunction with kernel functions. However, a common difficulty encountered when implementing kernel-based online algorithms is the amount of memory required to store the online hypothesis, which may grow unboundedly as the algorithm progresses. Moreover, the running time of each online round grows linearly with the amount of memory used to store the hypothesis. In this paper, we present the Forgetron family of kernel-based online classification algorithms, which overcome this problem by restricting themselves to a predefined memory budget. We obtain different members of this family by modifying the kernel-based Perceptron in various ways. We also prove a unified mistake bound for all of the Forgetron algorithms. To our knowledge, this is the first online kernel-based learning paradigm which, on one hand, maintains a strict limit on the amount of memory it uses and, on the other hand, entertains a relative mistake bound. We conclude with experiments using real datasets, which underscore the merits of our approach.

Extensions of vector quantization for incremental clustering

In this paper, we extend the conventional vector quantization by incorporating a vigilance parameter, which steers the tradeoff between plasticity and stability during incremental online learning. This is motivated in the adaptive resonance theory (ART) network approach and is exploited in our paper for forming a one-pass incremental and evolving variant of vector quantization. This variant can be applied for online clustering, classification and approximation tasks with an unknown number of clusters. Additionally, two novel extensions are described: one concerns the incorporation of the sphere of influence of clusters in the vector quantization learning process by selecting the ‘winning cluster’ based on the distances of a data point to the surface of all clusters. Another one introduces a deletion of cluster satellites and an online split-and-merge strategy: clusters are dynamically split and merged after each incremental learning step. Both strategies prevent the algorithm to generate a wrong cluster partition due to a bad a priori setting of the most essential parameter(s). The extensions will be applied to clustering of two- and high-dimensional data, within an image classification framework and for model-based fault detection based on data-driven evolving fuzzy models.