Parallel Programming and Performance Evaluation with the URSA Tool Family

This paper contributes to the solution of several open problems with parallel programming tools and their integration with performance evaluation environments. First, we propose interactive compilation scenarios instead of the usual black-box-oriented use of compiler tools. In such scenarios, information gathered by the compiler and the compiler's reasoning are presented to the user in meaningful ways and on-demand. Second, a tight integration of compilation and performance analysis tools is advocated. M any of the existing, advanced instruments for gathering performance results are being used in the presented environment and their results are combined in integrated views with compiler information and data from other tools. Initial instruments that assist users in data mining this information are presented and the need for much stronger facilities is explained. The URSA Family provides two tools addressing these issues. URSA MINOR supports a group of users at a specific site, such as a research or development project. URSA MAJOR complements this tool by making available the gathered results to the user community at large via the World-wide Web. This paper presents objectives, functionality, experience, and next development steps of the URSA tool family. Two case studies are presented that illustrate the use of the tools for developing and studying parallel applications and for evaluating parallelizing compilers.     

Detection of unanticipated faults for autonomous underwater vehicles using online topic models

For robots to succeed in complex missions, they must be reliable in the face of subsystem failures and environmental challenges. In this paper, we focus on autonomous underwater vehicle (AUV) autonomy as it pertains to self‐perception and health monitoring, and we argue that automatic classification of state‐sensor data represents an important enabling capability. We apply an online Bayesian nonparametric topic modeling technique to AUV sensor data in order to automatically characterize its performance patterns, then demonstrate how in combination with operator‐supplied semantic labels these patterns can be used for fault detection and diagnosis by means of a nearest‐neighbor classifier. The method is evaluated using data collected by the Monterey Bay Aquarium Research Institute's Tethys long‐range AUV in three separate field deployments. Our results show that the proposed method is able to accurately identify and characterize patterns that correspond to various states of the AUV, and classify faults at a high rate of correct detection with a very low false detection rate.     

Multi-dimensional confocal laser scanning microscopy image correlation for nanoparticle flow velocimetry

We present a new multi-dimensional confocal laser scanning microscopy (CLSM) image correlation for nanoparticle flow velocimetry that is robust to sources of decorrelating errors. Random and bias errors from nanoparticle flow measurements exacerbate with increased dimensionality in CLSM images, rendering measurements unusable. Our new algorithm tackles these measurement limitations in twofold. First, we model and correct for the bias errors introduced by the effects of the volumetric laser scanning image acquisition. Second, we developed a new spectral filter using a phase-quality masking technique that optimizes its size for the spectral content of CLSM images, without requiring a priori knowledge of displacement fields or flow tracer properties. We validated our algorithm using synthetic images and experimentally obtained 2D and 3D CLSM images of nanoparticle flow through a micro-channel. We show that our technique significantly outperforms the standard cross-correlation (SCC) in reducing both the random and bias errors and accelerated the convergence of ensemble correlation velocity measurements from CLSM images.     

Deterministic lateral displacement (DLD) in the high Reynolds number regime: high-throughput and dynamic separation characteristics

Recent progress in the development of biosensors has created a demand for high-throughput sample preparation techniques that can be easily integrated into microfluidic or lab-on-a-chip platforms. One mechanism that may satisfy this demand is deterministic lateral displacement (DLD), which uses hydrodynamic forces to separate particles based on size. Numerous medically relevant cellular organisms, such as circulating tumor cells (10–15 µm) and red blood cells (6–8 µm), can be manipulated using microscale DLD devices. In general, these often-viscous samples require some form of dilution or other treatment prior to microfluidic transport, further increasing the need for high-throughput operation to compensate for the increased sample volume. However, high-throughput DLD devices will require a high flow rate, leading to an increase in Reynolds numbers (Re) much higher than those covered by existing studies for microscale (≤?100 µm) DLD devices. This study characterizes the separation performance for microscale DLD devices in the high-Re regime (10?<?Re?<?60) through numerical simulation and experimental validation. As Re increases, streamlines evolve and microvortices emerge in the wake of the pillars, resulting in a particle trajectory shift within the DLD array. This differs from previous DLD works, in that traditional models only account for streamlines that are characteristic of low-Re flow, with no consideration for the transformation of these streamlines with increasing Re. We have established a trend through numerical modeling, which agrees with our experimental findings, to serve as a guideline for microscale DLD performance in the high-Re regime. Finally, this new phenomenon could be exploited to design passive DLD devices with a dynamic separation range, controlled simply by adjusting the device flow rate.     

Information from social networking sites: Context collapse and ambiguity in the hiring process

Forming impressions of job candidates is a challenging process, one characterized by ambiguity brought about by the uncertainty associated with making decisions and judgments. To reduce ambiguity, hiring professionals have established policies and procedures to facilitate the sourcing and use of information about a candidate. However, recently, a public source of information is increasingly being used—information from social networking sites (SNSs). While conventional wisdom says more information is better and can help make decisions less ambiguous, this relationship may not be as straightforward as expected when facing assessments of candidates. This paper examines two such aspects, information‐task quality and context collapse, and their collective impact on ambiguity when making an assessment of a job candidate. Using data from an online survey‐based experiment, the findings suggest information from SNSs can be useful, yet can create ambiguity for decision makers because of context collapse made possible by SNS technologies.     

Broadband Forward Light Scattering by Architectural Design of Core–Shell Silicon Particles

A goal in the field of nanoscale optics is the fabrication of nanostructures with strong directional light scattering at visible frequencies. Here, the synthesis of Mie-resonant core–shell particles with overlapping electric and magnetic dipole resonances in the visible spectrum is demonstrated. The core consists of silicon surrounded by a lower index silicon oxynitride (SiO_xN_y) shell of an adjustable thickness. Optical spectroscopies coupled to Mie theory calculations give the first experimental evidence that the relative position and intensity of the magnetic and electric dipole resonances are tuned by changing the core–shell architecture. Specifically, coating a high-index particle with a low-index shell coalesces the dipoles, while maintaining a high scattering efficiency, thus generating broadband forward scattering. This synthetic strategy opens a route toward metamaterial fabrication with unprecedented control over visible light manipulation.     

Rigid formation control of double-integrator systems

In this paper, we study rigid formation control systems modelled by double integrators. Two kinds of double-integrator formation systems are considered, namely formation stabilisation systems and flocking control systems. Novel observations on the measurement requirement, the null space and eigenvalues of the system Jacobian matrix will be provided, which reveal important properties of system dynamics and the associated convergence results. We also establish some new links between single-integrator formation systems and double-integrator formation systems via a parameterised Hamiltonian system, which, in addition, provide novel stability criteria for different equilibria in double-integrator formation systems by using available results in single-integrator formation systems.     

Salient Object Subitizing

We study the problem of salient object subitizing, i.e. predicting the existence and the number of salient objects in an image using holistic cues. This task is inspired by the ability of people to quickly and accurately identify the number of items within the subitizing range (1–4). To this end, we present a salient object subitizing image dataset of about 14 K everyday images which are annotated using an online crowdsourcing marketplace. We show that using an end-to-end trained convolutional neural network (CNN) model, we achieve prediction accuracy comparable to human performance in identifying images with zero or one salient object. For images with multiple salient objects, our model also provides significantly better than chance performance without requiring any localization process. Moreover, we propose a method to improve the training of the CNN subitizing model by leveraging synthetic images. In experiments, we demonstrate the accuracy and generalizability of our CNN subitizing model and its applications in salient object detection and image retrieval.