Multi-level graph layout on the GPU

This paper presents a new algorithm for force directed graph layout on the GPU. The algorithm, whose goal is to compute layouts accurately and quickly, has two contributions. The first contribution is proposing a general multi-level scheme, which is based on spectral partitioning. The second contribution is computing the layout on the GPU. Since the GPU requires a data parallel programming model, the challenge is devising a mapping of a naturally unstructured graph into a well-partitioned structured one. This is done by computing a balanced partitioning of a general graph. This algorithm provides a general multi-level scheme, which has the potential to be used not only for computation on the GPU, but also on emerging multi-core architectures. The algorithm manages to compute high quality layouts of large graphs in a fraction of the time required by existing algorithms of similar quality. An application for visualization of the topologies of ISP (Internet Service Provider) networks is presented.     

Concurrent Java Test Generation as a Search Problem

A Random test generator generates executable tests together with their expected results. In the form of a noise-maker, it seeds the program with conditional scheduling primitives (such as yield()) that may cause context switches. As a result different interleavings are potentially produced in different executions of the program. Determining a-priori the set of seeded locations required for a bug to manifest itself is rarely possible.This work proposes to reformulate random test generation of concurrent Java programs as a search problem. Hence, it allows applying a set of well known search techniques from the domain of AI to the input space of the test generator. By iteratively refining the input parameters fed to the test generator, the search process creates testing scenarios (i.e. interleavings) that maximizes predefined objective functions. We develop geneticFinder, a noise-maker that uses a genetic algorithm as a search method. We demonstrate our approach by maximizing two objective functions: the high manifestation rate of concurrent bugs and the exporting of a high degree of debugging information to the user. Experimental results show our approach is effective.     

Multiclass object classification for real-time video surveillance systems

Object classification in video is an important factor for improving the reliability of various automatic applications in video surveillance systems, as well as a fundamental feature for advanced applications, such as scene understanding. Despite extensive research, existing methods exhibit relatively moderate classification accuracy when tested on a large variety of real-world scenarios, or do not obey the real-time constraints of video surveillance systems. Moreover, their performance is further degraded in multi-class classification problems. We explore multi-class object classification for real-time video surveillance systems and propose an approach for classifying objects in both low and high resolution images (human height varies from a few to tens of pixels) in varied real-world scenarios. Firstly, we present several features that jointly leverage the distinction between various classes. Secondly, we provide a feature-selection procedure based on entropy gain, which screens out superfluous features. Experiments, using various classification techniques, were performed on a large and varied database consisting of ∼29,000 object instances extracted from 140 different real-world indoor and outdoor, near-field and far-field scenes having various camera viewpoints, which capture a large variety of object appearances under real-world environmental conditions. The insight raised from the experiments is threefold: the efficiency of our feature set in discriminating between classes, the performance improvement when using the feature selection method, and the high classification accuracy obtained on our real-time system on both DSP (TMS320C6415-6E3, 600 MHz) and PC (Quad Core Intel® Xeon® E5310, 2 × 4 MB Cache, 1.60 GHz, 1066 MHz) platforms.     

Multi-input/single-output computer-aided control design using the quantitative feedback theory

The quantitative feedback theory is an engineering design technique of uncertain feedback systems having robust stability and robust performance specifications. The crux of the quantitative feedback theory is a transformation of robust stability and robust performance specifications into domains in the complex plane, referred to as bounds, where a nominal loop transmission should lie within. To date, a quantitative feedback theory design is being carried out using manual (i.e. graphical) procedures or search algorithms. This paper shows that there exists a formal map from the uncertain plant and each closed-loop specification to these bounds. In particular, it is shown that each map has a closed form consisting of a quadratic inequality. These maps greatly simplify the computational aspects of the quantitative feedback theory in design of single-loop feedback systems. Based on this new development, a simple-to-implement, efficient computer algorithm is outlined.     

Evaluation Metrics for Conditional Image Generation

International Journal of Computer Vision - We present two new metrics for evaluating generative models in the class-conditional image generation setting. These metrics are obtained by generalizing...     

Silicon-Rich Higher Manganese Silicides for Thermoelectric Applications

Thermoelectrics are widely explored as a renewable and environmentally friendly method for converting waste heat to electrical power. Higher manganese silicides (HMS) have been identified as promising, nontoxic, inexpensive, p-type materials for thermoelectric applications. To mass-produce practical thermoelectric converters, an inexpensive, effective, and simple production method should be applied for these materials. In the frame of the current research, HMS have been synthesized using furnace melting followed by powdering and spark plasma sintering. Highly dense samples were obtained and measured for their thermoelectric properties. The samples were also characterized using x-ray diffraction, scanning electron microscopy, and Hall-effect measurements. Homogeneous samples were obtained with small inclusions of Si, reaching a figure of merit of about 0.6 at 450°C.     

Using the biophysical CHARMED model to elucidate the underpinnings of contrast in diffusional kurtosis analysis of diffusion-weighted MRI

Object

The aim of this work was to understand biophysical substrate underpinning contrast in diffusional kurtosis imaging (DKI) in white matter, using the composite hindered and restricted model of diffusion (CHARMED).

Materials and methods

A theoretical relationship between the kurtosis function K and the CHARMED parameters, i.e., the restricted volume fraction RF and the axonal longitudinal diffusivity D was derived for the propagator used in the CHARMED model. Evidence for a similar correlation between these measures was then investigated in vivo across different WM regions in five healthy young subjects that underwent a CHARMED protocol at 3T.

Results

Our theoretical treatment shows that K has an increasing trend for both increasing RF values and increasing D. In vivo, a significant positive correlation (P?K _⊥ and RF. A multilinear regression showed that K _⊥ values are better explained by a mixed contribution of both RF and D.

Conclusions

The CHARMED model was used to understand whether and where DKI contrast can be explained in terms of the underlying axonal geometry. This work demonstrates that the information contained in DKI overlaps with the information extracted by CHARMED in areas of higher intra-voxel directional coherence.