IMPROVING MULTILABEL CLASSIFICATION BY AVOIDING IMPLICIT NEGATIVITY WITH INCOMPLETE DATA

Many real‐world problems require multilabel classification, in which each training instance is associated with a set of labels. There are many existing learning algorithms for multilabel classification; however, these algorithms assume implicit negativity, where missing labels in the training data are automatically assumed to be negative. Additionally, many of the existing algorithms do not handle incremental learning in which new labels could be encountered later in the learning process. A novel multilabel adaptation of the backpropagation algorithm is proposed that does not assume implicit negativity. In addition, this algorithm can, using a naïve Bayesian approach, infer missing labels in the training data. This algorithm can also be trained incrementally as it dynamically considers new labels. This solution is compared with existing multilabel algorithms using data sets from multiple domains, and the performance is measured with standard multilabel evaluation metrics. It is shown that our algorithm improves classification performance for all metrics by an overall average of 7.4% when at least 40% of the labels are missing from the training data and improves by 18.4% when at least 90% of the labels are missing.     

Quality assessment of false‐colored fused displays

Abstract— The problem of assessing the quality of fused images (composites created from inputs of differing modalities, such as infrared and visible light radiation) is an important and growing area of research. Recent work has shown that the process of assessing fused images should not rely entirely on subjective quality methods, with objective tasks and computational metrics having important contributions to the assessment procedure. The current paper extends previous findings, applying a psychophysical selection task, metric evaluation, and subjective quality judgments to a range of fused surveillance images. Fusion schemes included the contrast pyramid and shift invariant discrete wavelet transform (Experiment 1), the complex wavelet transform (Experiments 1 & 2), and two false‐coloring methods (Experiment 2). In addition, JPEG2000 compression was applied at two levels, as well as an uncompressed control. Reaction time results showed the contrast pyramid to lead to slowest performance in the objective task, whilst the presence of color greatly reduced reaction times. These results differed from both the subjective and metric results. The findings support the view that subjective quality ratings should be used with caution, especially if not accompanied by some task.     

Guest Editors' Introduction: Strategic Planning in Today's IT Environment

This introduction to the special issue dedicated to strategic planning presents the complexity and dynamics of global enterprise IT management in today's tight and unstable economic environment. IT managers and enterprise executives can achieve their overall business success by opening channels of communications among their normally independent organizational silos and by creating an enterprise-shared culture that recognizes the critical role IT has in this endeavor.     

An efficient scheme of target classification and information fusion in wireless sensor networks

In this paper, an efficient target classification and fusion scheme for wireless sensor networks (WSNs) is proposed and evaluated. When a classification algorithm for WSN nodes is designed, parametric approaches such as Gaussian mixture model (GMM) should be more preferred to non-parametric ones due to the hard limitation in resources. The GMM algorithm not only shows good performances for target classification in WSNs but it also requires very small resources. Based on the classifier, a decision tree generated by the classification and regression tree algorithm is used to fuse the information from heterogeneous sensors. This node-level classification scheme provides a satisfactory classification rate, 94.10%, with little resources. Finally, a confidence-based fusion algorithm improves the overall accuracy by fusing the information among sensor nodes. Our experimental results show that the proposed group-level fusion algorithm improves the accuracy by an average of 4.17% accuracy with randomly selected nodes.     

Requirements elicitation for virtual actors in collaborative learning environments

This paper reports on the elicitation of requirements for Virtual Actors in Collaborative Virtual Learning Environments (CVLEs). The methodological approach followed involves the phased development of a series of learning environments which are observed in use by parents, children and teachers. The focus of study is on the interactivity and social communication issues that arise in the learning situation. The research uses as its case study the work of the Manchester Museum Education Service with children at Key Stage Level 2 (9–11 years old) of the National Curriculum. The particular learning situation is based on senet, an ancient Egyptian board game from the Museum’s collection of artefacts from the pyramid builders’ town of Kahun. Results are presented of the first phase prototype, a single display groupware system where interactions take place face-to-face in the ‘real-world’ external to the environment. Results are also presented of the second phase prototype, a multi-user groupware environment in which the users are remotely located and interaction is mainly internal to the environment. The paper discusses how the results from these two phases are being used to establish requirements for a CVLE to be developed in the third phase of research.     

Detection and removal of anomalies in digitized animation film

In this article, the problem of detecting and removing anomalies in digitized animation film is addressed. The impetus of this article comes from the motion picture industry, where several studios are rereleasing vintage film to the public which are often accompanied by visual degradation in the film color, grain, and overall quality. These degradations, or anomalies, can be seen as unwanted visual artifacts that are usually larger than one pixel in size and appear in only one frame. One of the novelties of this article is that the restoration of animation film, as opposed to “real-world” image sequences, is investigated. It often contains additional artifacts and raises other issues not found in most sequences. These artifacts are first defined, and the appropriate steps for their detection and removal are described. Experiments with the proposed algorithm were performed using scenes from the animation film Fantasia, courtesy of Walt Disney Feature Animations, and are discussed in detail. Although our algorithm was developed for the removal of artifacts on animation film, aspects of it can be applied to nonanimated film. © 1998 John Wiley & Sons, Inc. Int J Imaging Syst Technol, 9, 283–293, 1998     

Ultrahigh Carrier Mobility Achieved in Photoresponsive Hybrid Perovskite Films via Coupling with Single‐Walled Carbon Nanotubes

Organolead trihalide perovskites have drawn substantial interest for photovoltaic and optoelectronic applications due to their remarkable physical properties and low processing cost. However, perovskite thin films suffer from low carrier mobility as a result of their structural imperfections such as grain boundaries and pinholes, limiting their device performance and application potential. Here we demonstrate a simple and straightforward synthetic strategy based on coupling perovskite films with embedded single‐walled carbon nanotubes. We are able to significantly enhance the hole and electron mobilities of the perovskite film to record‐high values of 595.3 and 108.7 cm² V^?1 s^?1, respectively. Such a synergistic effect can be harnessed to construct ambipolar phototransistors with an ultrahigh detectivity of 3.7 × 10¹⁴ Jones and a responsivity of 1 × 10⁴ A W^?1, on a par with the best devices available to date. The perovskite/carbon nanotube hybrids should provide a platform that is highly desirable for fields as diverse as optoelectronics, solar energy conversion, and molecular sensing.     

Electron Transfer Mediated by Surface‐Tethered Redox Groups in Nanofluidic Devices

Electrochemistry provides a powerful sensor transduction and amplification mechanism that is highly suited for use in integrated, massively parallelized assays. Here, the cyclic voltammetric detection of flexible, linear poly(ethylene glycol) polymers is demonstrated, which have been functionalized with redox‐active ferrocene (Fc) moieties and surface‐tethered inside a nanofluidic device consisting of two microscale electrodes separated by a gap of <100 nm. Diffusion of the surface‐bound polymer chains in the aqueous electrolyte allows the redox groups to repeatedly shuttle electrons from one electrode to the other, resulting in a greatly amplified steady‐state electrical current. Variation of the polymer length provides control over the current, as the activity per Fc moiety appears to depend on the extent to which the polymer layers of the opposing electrodes can interpenetrate each other and thus exchange electrons. These results outline the design rules for sensing devices that are based on changing the polymer length, flexibility, and/or diffusivity by binding an analyte to the polymer chain. Such a nanofluidic enabled configuration provides an amplified and highly sensitive alternative to other electrochemical detection mechanisms.     

Shape‐Enhanced Photocatalytic Activities of Thoroughly Mesoporous ZnO Nanofibers

1D mesoporous materials have attracted extensive interest recently, owning to their fascinating properties and versatile applications. However, it remains as a grand challenge to develop a simple and efficient technique to produce oxide nanofibers with mesoporous architectures, controlled morphologies, large surface areas, and optimal performances. In this work, a facile foaming‐assisted electrospinning strategy with foaming agent of tea saponin is used to produce thoroughly mesoporous ZnO nanofibers with high purity and controlled morphology. Interestingly, mesoporous fibers with elliptical cross‐section exhibit the significantly enhanced photocatalytic activity for hydrogen production, as compared to the counterparts with circular and rectangular cross‐sections, and they also perform better than the commercial ZnO nanopowders. The unexpected shape dependence of photocatalytic activities is attributed to the different stacking modes of the mesoporous fibers, and a geometrical model is developed to account for the shape dependence. This work represents an important step toward producing thoroughly mesoporous ZnO nanofibers with tailored morphologies, and the discovery that fibers with elliptical cross‐section render the best performance provides a valuable guideline for improving the photocatalytic performance of such mesoporous nanomaterials.