"... In PAGE 4: ... First, it resizes the images based on the device profile. Table1 shows a resultant data from a test session with Microsoft Power point. The resolution is customizable on the client side.... In PAGE 4: ...etwork bandwidth. Garnet compresses image data with Huffman [35] and LZ77 [36] algorithm. For raw bitmap data, comprising of 8 bit of RGB and alpha, Garnet eliminates alpha bits which represents transparency of image, and reduces the communicating data size to 25%. Table1 presents the actual data size transformed via a wireless ... In PAGE 5: ... Performing transcoding in proxy-based middleware may have disadvantages in a large user group. As we presented in Table1 , the actual data delivered to mobile devices are much smaller than the original data from master client. However, the transcoding functions should download whole amount of data to process it, which can incur significant (a) Bitmap Image (b)Vector Graphics (a) (b) Figure 2.... ..."

"... In PAGE 8: ... 5.2 Proximity to the shared display Interacting while close to the shared display was perceived by a majority of participants as being more effective (17/24) and enjoyable (16/24) for map interactions (see Table5 for combined rankings according to distance from the display). Superimposed pen-based interactions were natural for participants accustomed to whiteboard interactions.... In PAGE 8: ... There are significant interaction benefits to using direct input with a secondary display since it enables users to interact up-close, leveraging their experience with desktop environments. One third of the twelve participants who used direct input when interacting at a distance ranked that condition as most enjoyable and effective overall for interacting with the main display (see Table5 ). The text annotations on the map were smaller and more legible when participants utilized direct input at a distance rather than the indirect input device or direct input on the large shared display.... ..."

"... In PAGE 9: ...Canonicalization. We apply the canonicalization rules to the data at the preprocessing stage and the results of classi- fication accuracy are shown in Table2 . It can be seen that the difference between the results with and without canoni- calization is insubstantial.... In PAGE 9: ... The cross-validation results of 2-gram and 4-gram for the data are shown in Table 3. Compared to the 1-gram results in Table2 the higher-grams do not provide any im- provement. Comparatively, the work on intrusion detection using sys- tem call sequences reported patterns with length greater than 1 give useful results [16].... ..."

"... In PAGE 9: ... Next we study some higher n-gram classi- fiers which capture the local order of events in different degrees. The cross-validation results of 2-gram and 4-gram for the data are shown in Table3 . Compared to the 1-gram results in Table 2 the higher-grams do not provide any im- provement.... ..."