Graft polymerization of vinyl acetate onto silica

The free‐radical graft polymerization of vinyl acetate onto nonporous silica particles was studied experimentally. The grafting procedure consisted of surface activation with vinyltrimethoxysilane, followed by free‐radical graft polymerization of vinyl acetate in ethyl acetate with 2,2′‐azobis(2,4‐dimethylpentanenitrile) initiator. Initial monomer concentration was varied from 10 to 40% by volume and the reaction was spanned from 50 to 70°C. The resulting grafted polymer, which was stable over a wide range of pH levels, consisted of polymer chains that are terminally and covalently bonded to the silica substrate. The experimental polymerization rate order, with respect to monomer concentration, ranged from 1.61 to 2.00, consistent with the kinetic order for the high polymerization regime. The corresponding rate order for polymer grafting varied from 1.24 to 1.43. The polymer graft yield increased with both initial monomer concentration and reaction temperature, and the polymer‐grafted surface became more hydrophobic with increasing polymer graft yield. The present study suggests that a denser grafted polymer phase of shorter chains was created upon increasing temperature. On the other hand, both polymer chain length and polymer graft density increased with initial monomer concentration. Atomic force microscopy–determined topology of the polymer‐grafted surface revealed a distribution of surface clusters and surface elevations consistent with the expected broad molecular‐weight distribution for free‐radical polymerization. © 2002 Wiley Periodicals, Inc. J Appl Polym Sci 87: 300–310, 2003     

Differentiating seawater and groundwater sulfate attack in Portland cement mortars

The study reported in this article deals with understanding the physical, chemical and microstructural differences in sulfate attack from seawater and groundwater. Portland cement mortars were completely immersed in solutions of seawater and groundwater. Physical properties such as length, mass, and compressive strength were monitored periodically. Thermal analysis was used to study the relative amounts of phases such as ettringite, gypsum, and calcium hydroxide, and microstructural studies were conducted by scanning electron microscopy. Portland cement mortars performed better in seawater solution compared to groundwater solution. The difference in performance could be attributed to the reduction in the quantity of the expansive attack products (gypsum and ettringite). The high Cl concentration of seawater could have played an important role by binding the C₃A to form chloroaluminate compounds, such as Friedel's salt (detected in the microstructural studies), and also by lowering the expansive potential of ettringite. Furthermore, the thicker layer of brucite forming on the specimens in seawater could have afforded better protection against ingress of the solution than in groundwater.     

Specimen edge effects on bending fatigue of carburized steel

The effects of specimen geometry on the fatigue behavior of SAE 4320 steel carburized at 927 °C were evaluated with two sets of cantilever bend specimens, one set machined with square edges and one set machined with round edges. The specimens with square edges exhibited a 13% lower fatigue limit. In comparison to the rounded samples, the lower fatigue limit in the square-edged samples was attributed to the presence of a higher volume fraction of retained austenite in the sample corners and a lower surface residual compressive stress. As a result of the differences in residual stress, preferential crack initiation sites existed in the square-edged samples at a location approximately 200 to 900 ώm from the square edge. The implications of this study on laboratory analyses of the bending fatigue performance of carburized gear steels are discussed.     

Distilled Collections from Textual Image Queries

We present a distillation algorithm which operates on a large, unstructured, and noisy collection of internet images returned from an online object query. We introduce the notion of a distilled set, which is a clean, coherent, and structured subset of inlier images. In addition, the object of interest is properly segmented out throughout the distilled set. Our approach is unsupervised, built on a novel clustering scheme, and solves the distillation and object segmentation problems simultaneously. In essence, instead of distilling the collection of images, we distill a collection of loosely cutout foreground “shapes”, which may or may not contain the queried object. Our key observation, which motivated our clustering scheme, is that outlier shapes are expected to be random in nature, whereas, inlier shapes, which do tightly enclose the object of interest, tend to be well supported by similar shapes captured in similar views. We analyze the commonalities among candidate foreground segments, without aiming to analyze their semantics, but simply by clustering similar shapes and considering only the most significant clusters representing non‐trivial shapes. We show that when tuned conservatively, our distillation algorithm is able to extract a near perfect subset of true inliers. Furthermore, we show that our technique scales well in the sense that the precision rate remains high, as the collection grows. We demonstrate the utility of our distillation results with a number of interesting graphics applications.     

Fusing distributional and experiential information for measuring semantic relatedness

Models of semantic relatedness have usually focused on language-based distributional information without taking into account “experiential data” concerning the embodied sensorial source of the represented concepts. In this paper, we present an integrative cognitive model of semantic relatedness. The model – semantic family resemblance – uses a variation of the co-product as a mathematical structure that guides the fusion of distributional and experiential information. Our algorithm provides superior results in a set expansion task and a significant correlation with two benchmarks of human rated word-pair similarity datasets.     

Improving trust modeling through the limit of advisor network size and use of referrals

This paper explores potential improvements to the trust modeling of agents in multi-agent systems when a social network of advisors is employed as part of the trust modeling, and in particular, examines means of optimizing the number of advisors that should be maintained in the social network. We propose three such improvements, two directly relating to the limit of advisor network size by either setting a maximum size for a buyer’s advisor network or setting a minimum trustworthiness threshold for agents to be accepted into that advisor network, and a third which uses an advisor referral system in combination with one of the first two network-limiting techniques. We provide experimental results in defence of our approach for two distinct trust modeling systems, and show how these optimizations can improve, sometimes significantly, the accuracy of different trust models (in the context of electronic marketplaces). We believe that the proposed techniques will be very useful for trust researchers seeking to improve the accuracy of their own trust models by setting the size and composition of advisor networks.     

Hardware Technologies for High-Performance Data-Intensive Computing

Data-intensive problems challenge conventional computing architectures with demanding CPU, memory, and I/O requirements. Experiments with three benchmarks suggest that emerging hardware technologies can significantly boost performance of a wide range of applications by increasing compute cycles and bandwidth and reducing latency.     

A framework for simulating real-time multi-agent systems

In this paper, we describe an implementation of use in demonstrating the effectiveness of architectures for real-time multi-agent systems. The implementation provides a simulation of a simplified RoboCup Search and Rescue environment, with unexpected events, and includes a simulator for both a real-time operating system and a CPU. We present experimental evidence to demonstrate the benefit of the implementation in the context of a particular hybrid architecture for multi-agent systems that allows certain agents to remain fully autonomous, while others are fully controlled by a coordinating agent. In addition, we discuss the value of the implementation for testing any models for the construction of real-time multi-agent systems and include a comparison to related work.

Multi-Class Segmentation with Relative Location Prior

Multi-class image segmentation has made significant advances in recent years through the combination of local and global features. One important type of global feature is that of inter-class spatial relationships. For example, identifying “tree” pixels indicates that pixels above and to the sides are more likely to be “sky” whereas pixels below are more likely to be “grass.” Incorporating such global information across the entire image and between all classes is a computational challenge as it is image-dependent, and hence, cannot be precomputed. In this work we propose a method for capturing global information from inter-class spatial relationships and encoding it as a local feature. We employ a two-stage classification process to label all image pixels. First, we generate predictions which are used to compute a local relative location feature from learned relative location maps. In the second stage, we combine this with appearance-based features to provide a final segmentation. We compare our results to recent published results on several multi-class image segmentation databases and show that the incorporation of relative location information allows us to significantly outperform the current state-of-the-art.     

Evaluation of Face Datasets as Tools for Assessing the Performance of Face Recognition Methods

Face datasets are considered a primary tool for evaluating the efficacy of face recognition methods. Here we show that in many of the commonly used face datasets, face images can be recognized accurately at a rate significantly higher than random even when no face, hair or clothes features appear in the image. The experiments were done by cutting a small background area from each face image, so that each face dataset provided a new image dataset which included only seemingly blank images. Then, an image classification method was used in order to check the classification accuracy. Experimental results show that the classification accuracy ranged between 13.5% (color FERET) to 99% (YaleB). These results indicate that the performance of face recognition methods measured using face image datasets may be biased. Compilable source code used for this experiment is freely available for download via the Internet.