China
Africa
Explore chapters and articles related to this topic
Increasing Projector Contrast and Brightness through Light Redirection
Published in Laurent A. Francis, Krzysztof Iniewski, Novel Advances in Microsystems Technologies and Their Applications, 2017
Reynald Hoskinson, Boris Stoeber
Our approach for allocating the light from all AMA mirrors is to divide the original image into equal energy zones and allocate one ML for each zone. The median cut algorithm described in Debevec (2005) is an efficient method of subdividing an image into zones of approximately equal energy. We have modified the original algorithm so that it can divide an image into an arbitrary number of regions d. The image is first added to the region list as a single region, along with the number of desired divisions. As long as d > 2, subdivide the region into two parts as equally as possible, adding the two new subregions back to the region list, as well as the new divided d for each. In order to maintain the constraint that all regions at the final level be of approximately equal energy, choose the cutting line so that it divides the region into portions of energy as equal as possible, along the largest dimension of the parent region. Repeat until all d’s in the region list equal 1.
Effect of hollow bit local exhaust ventilation on respirable quartz dust concentrations during concrete drilling
Published in Journal of Occupational and Environmental Hygiene, 2019
David Rempel, Alan Barr, Michael R. Cooper
Respirable silica dust sampling followed ISO and NIOSH standards. A “sampling” mannequin was fixed behind the drill in a location similar to where a worker would be. During each trial, a respirable dust sample (4 µm median cut point) was collected within 30 cm of the mannequin’s nose or mouth. The sampler was a GK 4.162 cyclone (BGI by Mesa Labs, Inc., Butler, NJ) holding a pre-weighed 47 mm polyvinyl chloride filter. A portable battery-powered pump (model SG10–2; GSA Messgerätebau GmbH, Neuss, Germany) was used to draw air through the sampler and was calibrated to provide a flow rate of 9 L min−1. The pump was calibrated before each session and verified after each sampling session with a digital volumetric flow meter (model 4146 primary calibrator; TSI Inc., Shoreview, MN).
Very Fast C4.5 Decision Tree Algorithm
Published in Applied Artificial Intelligence, 2018
Anis Cherfi, Kaouther Nouira, Ahmed Ferchichi
In accordance with Table 2 and Figure 1, the cut points provided by the mean is slightly higher than the median cut point which is explained by the presence of extreme values in both scenario 4 and 7. Thus, in those cases, median cut points are more accurate than the mean ones. On the other hand, the mean cut point shows a smallest increase in error rate when instances are strongly fluctuated (i.e., scenario 5). Otherwise, a closer inspection of the results in Table 2 shows that in all scenarios there is at least one of the two cut points generated by mean and median that is slightly better or close to the C4.5 threshold cut point results. Thus, it is necessary to adopt both of mean and median as a candidate cut points and then to select the one that maximizes information gain. As shown in Figure 2, the associated sensitivities and specificities found by mean and median method are usually slightly higher than those detected with the C4.5 threshold technique. As already shown in Table 2 for the mean and median cut points, also for sensitivities and specificities from Figure 2, using both the mean and median as cut points can yield to slightly higher or the same results as obtained using the C4.5 threshold technique.