A scalable kernel-based semisupervised metric learning algorithm with out-of-sample generalization ability

In recent years, metric learning in the semisupervised setting has aroused a lot of research interest. One type of semisupervised metric learning utilizes supervisory information in the form of pairwise similarity or dissimilarity constraints. However, most methods proposed so far are either limited to linear metric learning or unable to scale well with the data set size. In this letter, we propose a nonlinear metric learning method based on the kernel approach. By applying low-rank approximation to the kernel matrix, our method can handle significantly larger data sets. Moreover, our low-rank approximation scheme can naturally lead to out-of-sample generalization. Experiments performed on both artificial and real-world data show very promising results.     

Continuous State Dynamic Programming via Nonexpansive Approximation

This paper studies fitted value iteration for continuous state numerical dynamic programming using nonexpansive function approximators. A number of approximation schemes are discussed. The main contribution is to provide error bounds for approximate optimal policies generated by the value iteration algorithm.      

Direct microscale imprinting of Al at room temperature with Si inserts

For direct imprinting of metals, hard materials such as diamond and SiC have been used to construct mold inserts in preference to Si, despite the ease in fabricating Si-based micro-/nano- scale structures. In this communication, we demonstrate that micron scale Al structures can be replicated with good fidelity at room temperature by compression molding with Si inserts without incurring insert damage. We further report on results of a finite element analysis of the mechanics of the molding process. The finite element results provide some understanding of the observed lack of damage to the Si inserts.     

Perceptual image preview

Image preview is a convenient way to browse large or multiple images on small displays. However, current signal-level image resampling algorithms may remove many features of interest in the preview image. In this paper, we propose perceptual image preview which retains more perceptual features such that users can inspect features of interest by viewing the preview image only and without zooming in. This technology has two components, structure enhancement and perceptual feature visualization. Structure enhancement enhances the image structure while suppressing subtle details using a gradient modulation method, thus making the succedent perceptual features more apparent. For perceptual feature visualization, features of interest detected in the picture is visualized on the structure enhanced preview image. We demonstrate with two examples of most commonly used image quality features, image blur and noise. The effectiveness of the proposed method is validated by experimental results.     

Re-examining the causal structure of information technology impact research

About 20 years ago, Markus and Robey noted that most research on IT impacts had been guided by deterministic perspectives and had neglected to use an emergent perspective, which could account for contradictory findings. They further observed that most research in this area had been carried out using variance theories at the expense of process theories. Finally, they suggested that more emphasis on multilevel theory building would likely improve empirical reliability. In this paper, we reiterate the observations and suggestions made by Markus and Robey on the causal structure of IT impact theories and carry out an analysis of empirical research published in four major IS journals, Management Information Systems Quarterly (MISQ), Information Systems Research (ISR), the European Journal of Information Systems (EJIS), and Information and Organization (I&O), to assess compliance with those recommendations. Our final sample consisted of 161 theory-driven articles, accounting for approximately 21% of all the empirical articles published in these journals. Our results first reveal that 91% of the studies in MISQ, ISR, and EJIS focused on deterministic theories, while 63% of those in I&O adopted an emergent perspective. Furthermore, 91% of the articles in MISQ, ISR, and EJIS adopted a variance model; this compares with 71% from I&O that applied a process model. Lastly, mixed levels of analysis were found in 14% of all the surveyed articles. Implications of these findings for future research are discussed.     

An ant direction hybrid differential evolution heuristic for the large-scale passive harmonic filters planning problem

This article presents a heuristic which combines the orthogonal array experiment technique and an ant direction hybrid differential evolution algorithm (ADHDEOA) for planning of a large-scale passive harmonic filters problem. The addressed problem has a multi-bus and under abundant harmonic current sources in the system. In this study, an orthogonal array is first conducted to obtain the initial solution. Next, an ant direction hybrid differential evolution (ADHDE) is applied to search for a near optimum solution. The objective is to minimize the cost of the filter, the total demand distortion of harmonic currents and the total harmonic distortion of voltages at each bus simultaneously. In order to determine a set of weights of each term in the objective function, the simplest and most efficient form of triangular membership functions has been considered. The designed heuristic ADHDEOA is applied to a practical harmonic problem in a steel plant, and three design schemes are compared to demonstrate the performance of the ADHDEOA.     

Stability and control of nonlinear systems described by retarded functional equations: a review of recent results

This paper reports on recent results in a series of the work of the authors on the stability and nonlinear control for general dynamical systems described by retarded functional differential and difference equations. Both internal and external stability properties are studied. The corresponding Lyapunov and Razuminkhin characterizations for input-to-state and input-to-output stabilities are proposed. Necessary and sufficient Lyapunov-like conditions are derived for robust nonlinear stabilization. In particular, an explicit controller design procedure is developed for a new class of nonlinear time-delay systems. Lastly, sufficient assumptions, including a small-gain condition, are presented for guaranteeing the input-to-output stability of coupled systems comprised of retarded functional differential and difference equations.     

Polarization discrimination between repeater false-target and radar target

High fidelity repeater false-target badly affects a radar system’s detecting, tracking, and data processing. It is an available approach of confronting false-target for radar that discriminates firstly and then eliminates. Whereas for the technique progress about the repeater false-target jam, it is more and more difficult to discriminate this jam in the time-domain, frequency-domain, or space-domain. The technique using polarization information to discriminate the target and false-target is discussed in th...     

The development of a sub-population genetic algorithm II (SPGA II) for multi-objective combinatorial problems

Previous research has shown that sub-population genetic algorithm is effective in solving the multi-objective combinatorial problems. Based on these pioneering efforts, this paper extends the SPGA algorithm with a global Pareto archive technique and a two-stage approach to solve the multi-objective problems. In the first stage, the areas next to the two single objectives are searched and solutions explored around these two extreme areas are reserved in the global archive for later evolutions. Then, in the second stage, larger searching areas except the middle area are further extended to explore the solution space in finding the near-optimal frontiers. Through extensive experimental results, SPGA II does outperform SPGA, NSGA II, and SPEA 2 in the parallel scheduling problems and knapsack problems; it shows that the approach improves the sub-population genetic algorithm significantly. It may be of interests for researchers in solving multi-objective combinatorial problems.