Efficient dimensionality reduction and one-class classification for content-based image retrieval

The thesis investigates various machine learning approaches to reducing data dimensionality, and studies the impact of asymmetric data on learning in image retrieval. Efficient algorithms are proposed to reduce the data dimensionality. Integration strategies for one-class classification are designed to address asymmetric data issue and improve retrieval effectiveness.

Variables associated with children's physical activity levels during recess : the A-CLASS project

Background
School recess provides a daily opportunity for children to engage in physically active behaviours. However, few studies have investigated what factors may influence children's physical activity levels in this context. Such information may be important in the development and implementation of recess interventions. The aim of this study was to investigate the association between a range of recess variables and children's sedentary, moderate and vigorous physical activity in this context.

Methods
One hundred and twenty-eight children (39% boys) aged 9-10 years old from 8 elementary schools had their physical activity levels observed during school recess using the System for Observing Children's Activity and Relationships during Play (SOCARP). Playground variables data were also collected at this time. Multilevel prediction models identified variables that were significantly associated with children's sedentary, moderate and vigorous physical activity during recess.

Results
Girls engaged in 13.8% more sedentary activity and 8.2% less vigorous activity than boys during recess. Children with no equipment provision during recess engaged in more sedentary activity and less moderate activity than children provided with equipment. In addition, as play space per child increased, sedentary activity decreased and vigorous activity increased. Temperature was a significant negatively associated with vigorous activity.

Conclusions
Modifiable and unmodifiable factors were associated with children's sedentary, moderate and vigorous physical activity during recess. Providing portable equipment and specifying areas for activities that dominate the elementary school playground during recess may be two approaches to increase recess physical activity levels, though further research is needed to evaluate the short and long-term impact of such strategies.

Fast food restaurant locations : could they be 'supersizing' local communities?

This review examines the current state of knowledge related to the location of fast food restaurants. Previous studies have revealed some communities are more exposed to fast food restaurants. However, the influence of increased exposure to fast food on dietary behaviours remains unresolved. This is identified as an area of priority for future research.

Examining influences on boys' and girls' physical activity patterns: the a-class project

Understanding children’s physical activity (PA) patterns and the factors that may influence PA are important for developing interventions within this population. One hundred and ten children aged 9–0 years from 8 schools had their PA patterns assessed over 7 days. Physiological and self-report data were also collected. Multilevel analyses revealed that cardiorespiratory fitness was a consistent, significant and positive predictor of weekday and weekend PA, while the availability of home sedentary activities was a significant but negative predictor of PA. Since a range of variables were associated with PA levels, intervention developers should be cognizant of variables that may influence children’s activity

Fast finite-time consensus of a class of high-order uncertain nonlinear systems

This paper poses and solves a new problem of consensus control where the task is to make the fixed-topology multi-agent network, with each agent described by an uncertain nonlinear system in chained form, to reach consensus in a fast finite time. Our development starts with a set of new sliding mode surfaces. It is proven that, on these sliding mode surfaces, consensus can be achieved if the communication graph has the proposed directed spanning tree. Next, we introduce the multi-surface sliding mode control to drive the sliding variables to the sliding mode surfaces in a fast finite time. The control Lyapunov function for fast finite time stability, motivated by the fast terminal sliding mode control, is used to prove the reachability of the sliding mode surface. A recursive design procedure is provided, which guarantees the boundedness of the control input.