Boosting the margin: A new explanation for the effectiveness of voting methods

One of the surprising recurring phenomena observed in experiments with boosting is that the test error of the generated classifier usually does not increase as its size becomes very large, and often is observed to decrease even after the training error reaches zero. In this paper, we show that this phenomenon is related to the distribution of margins of the training examples with respect to the generated voting classification rule, where the margin of an example is simply the difference between the number of correct votes and the maximum number of votes received by any incorrect label. We show that techniques used in the analysis of Vapnik's support vector classifiers and of neural networks with small weights can be applied to voting methods to relate the margin distribution to the test error. We also show theoretically and experimentally that boosting is especially effective at increasing the margins of the training examples. Finally, we compare our explanation to those based on the bias-variance decomposition.

Heat strain, hydration status, and symptoms of heat illness in surface mine workers

The aim of the research program was to evaluate the heat strain, hydration status, and heat illness symptoms experienced by surface mine workers. An initial investigation involved 91 surface miners completing a heat stress questionnaire; assessing the work environment, hydration practices, and heat illness symptom experience. The key findings included 1) more than 80 % of workers experienced at least one symptom of heat illness over a 12 month period; and 2) the risk of moderate symptoms of heat illness increased with the severity of dehydration. These findings highlight a health and safety concern for surface miners, as experiencing symptoms of heat illness is an indication that the physiological systems of the body may be struggling to meet the demands of thermoregulation. To illuminate these findings a field investigation to monitor the heat strain and hydration status of surface miners was proposed. Two preliminary studies were conducted to ensure accurate and reliable data collection techniques. Firstly, a study was undertaken to determine a calibration procedure to ensure the accuracy of core body temperature measurement via an ingestible sensor. A water bath was heated to several temperatures between 23 . 51 ¢ªC, allowing for comparison of the temperature recorded by the sensors and a traceable thermometer. A positive systematic bias was observed and indicated a need for calibration. It was concluded that a linear regression should be developed for each sensor prior to ingestion, allowing for a correction to be applied to the raw data. Secondly, hydration status was to be assessed through urine specific gravity measurement. It was foreseeable that practical limitations on mine sites would delay the time between urine collection and analysis. A study was undertaken to assess the reliability of urine analysis over time. Measurement of urine specific gravity was found to be reliable up to 24 hours post urine collection and was suitable to be used in the field study. Twenty-nine surface miners (14 drillers [winter] and 15 blast crew [summer]) were monitored during a normal work shift. Core body temperature was recorded continuously. Average mean core body temperature was 37.5 and 37.4 ¢ªC for blast crew and drillers, with average maximum body temperatures of 38.0 and 37.9 ¢ªC respectively. The highest body temperature recorded was 38.4 ¢ªC. Urine samples were collected at each void for specific gravity measurement. The average mean urine specific gravity was 1.024 and 1.021 for blast crew and drillers respectively. The Heat Illness Symptoms Index was used to evaluate the experience of heat illness symptoms on shift. Over 70 % of drillers and over 80 % of blast crew reported at least one symptom. It was concluded that 1) heat strain remained within the recommended limits for acclimatised workers; and 2) the majority of workers were dehydrated before commencing their shift, and tend to remain dehydrated for the duration. Dehydration was identified as the primary issue for surface miners working in the heat. Therefore continued study focused on investigating a novel approach to monitoring hydration status. The final aim of this research program was to investigate the influence dehydration has on intraocular pressure (IOP); and subsequently, whether IOP could provide a novel indicator of hydration status. Seven males completed 90 minutes of walking in both a cool and hot climate with fluid restriction. Hydration variables and intraocular pressure were measured at baseline and at 30 minute intervals. Participants became dehydrated during the trial in the heat but maintained hydration status in the cool. Intraocular pressure progressively declined in the trial in the heat but remained relatively stable when hydration was maintained. A significant relationship was observed between intraocular pressure and both body mass loss and plasma osmolality. This evidence suggests that intraocular pressure is influenced by changes in hydration status. Further research is required to determine if intraocular pressure could be utilised as an indirect indicator of hydration status.