Manifold Based Local Classifiers: Linear and Nonlinear Approaches

In case of insufficient data samples in high-dimensional classification problems, sparse scatters of samples tend to have many ‘holes’—regions that have few or no nearby training samples from the class. When such regions lie close to inter-class boundaries, the nearest neighbors of a query may lie in the wrong class, thus leading to errors in the Nearest Neighbor classification rule. The K-local hyperplane distance nearest neighbor (HKNN) algorithm tackles this problem by approximating each class with a smooth nonlinear manifold, which is considered to be locally linear. The method takes advantage of the local linearity assumption by using the distances from a query sample to the affine hulls of query’s nearest neighbors for decision making. However, HKNN is limited to using the Euclidean distance metric, which is a significant limitation in practice. In this paper we reformulate HKNN in terms of subspaces, and propose a variant, the Local Discriminative Common Vector (LDCV) method, that is more suitable for classification tasks where the classes have similar intra-class variations. We then extend both methods to the nonlinear case by mapping the nearest neighbors into a higher-dimensional space where the linear manifolds are constructed. This procedure allows us to use a wide variety of distance functions in the process, while computing distances between the query sample and the nonlinear manifolds remains straightforward owing to the linear nature of the manifolds in the mapped space. We tested the proposed methods on several classification tasks, obtaining better results than both the Support Vector Machines (SVMs) and their local counterpart SVM-KNN on the USPS and Image segmentation databases, and outperforming the local SVM-KNN on the Caltech visual recognition database.     

Chemical vapour transport of transition metal oxides (III) crystal growth of Ti1−xRuxO2 and doped TiO2

Single crystals of the system Ti_1?xRu_xO₂ have been prepared by chemical vapour transport in closed ampoules using halide transport agents. The crystals are single phased for x ≦ 0.02 and 0.98 ≦ x. In the intermediate range a highly ordered two phased system is found. The pure RuO₂ crystals are substantially larger than any grown by open sysem chemical vapour transport. Single crystals of TiO₂ doped with other impurities have also been prepared.     

Mossbauer evidence for Ru(IV) and Ru(II) in TiO2

The Mössbauer absorption of the 89.4 keV line of ⁹⁹Ru has been measured at 4.2 K for samples of Ru in TiO₂ and in TiO₂ doped with Ta. We deduce that Ru replaces Ti substitutionally in TiO₂ as Ru(IV) and that no delocalisation of the four Ru(IV) 4d electrons occurs. If the donor Ta is added to the system, the ion Ru(II) is formed, demonstrating that Ru behaves as an acceptor in TiO₂.     

Detection of emergency vehicles: driver responses to advance warning in a driving simulator

OBJECTIVE: This research evaluated the effects of an advance warning device (AWD) on the safety of driver interactions with emergency vehicles (EVs). The AWD was intended to provide drivers with advance warning of an approaching on-call EV via visual and auditory warnings when the EV was within a 300- to 400-m radius. BACKGROUND: Research suggests that drivers can experience difficulty accurately detecting the distance and direction of approaching on-call EV. In-vehicle technology has not previously been explored as a means of overcoming the limitations of existing EV lights and sirens and improving driver detection of EV. METHOD: An experimental study using an advanced driving simulator examined the effects of the AWD on driving performance in a range of circumstances in which real-world EV crashes and near-misses commonly occur. Each event contained a combination of scenario type (adjacent lane, turning across, car following) and warning condition (control, standard, advance). RESULTS: Data from 22 participants were collected, including measures of speed, braking, and visual scanning. For adjacent-lane and turning-across events, the AWD was associated primarily with reductions in mean speed. The AWD resulted in an earlier lane change to clear a path for the EV in the car-following event. CONCLUSION: The reduction in speed observed was a positive finding, given the relationship between impact speed and injury severity. Response priming emerged as the mechanism underpinning these effects. Application: Response priming may result in safety benefits in other settings when an advisory warning is presented before the threat can be perceived.     

Transforming teaching and learning: embedding ICT into everyday classroom practices

Drawing on socio‐cultural theory, this paper describes how teams of teachers and researchers have developed ways of embedding information and communications technology (ICT) into everyday classroom practices to enhance learning. The focus is on teaching and learning across a range of subjects: English, history, geography, mathematics, modern foreign languages, music and science. The influence of young people's out‐of‐school uses of ICT on in‐school learning is discussed. The creative tension between idiosyncratic and institutional knowledge construction is emphasised and we argue that this is exacerbated by the use of ICT in the classroom.     

Visual accommodation and target detection in the vicinity of a window post

The visual environment of a driver in a car or a pilot in an airplane has intervening objects from the vehicle such as A-pillar posts between the windscreen and the doors in the car or cockpit pillars in the airplane. The presence of such objects can bias the observer's visual accommodation response because of the Mandelbaum effect (e.g., Owens, 1979). When subjects were allowed to focus toward a distance by looking through a large aperture in an intervening post, the detection (monocular) of a briefly presented distant target was found to be significantly better than when no aperture was present. When the size of the aperture was decreased from 2.3 to 1.15 deg diameter, target detection performance was significantly decreased and remained constant as further reduction of foveal cues was made. Although the detection results were generally in agreement with the visual accommodation results, detection accuracy changed significantly only with marked changes in accommodation. In addition to an accommodation bias, interference to target detection was also observed for those targets occurring at a laterally proximal position to the intervening object.