Results 1 - 10 of 54,552

Table 4. Number of selected features by feature selection methods in binary case

in Experimental Comparison of Feature Subset Selection Methods
by Chulmin Yun, Jihoon Yang
"... In PAGE 5: ... We conducted the same experiments with those datasets. As expected, all methods reduced the number of features ( Table4 ). However, the learning performance with selected features did not improve in all methods.... ..."

Table 2: Binary Scheduling Wheel Selection Process

in Design of a weighted fair queueing cell scheduler for ATM networks
by Yuhua Chen, Jonathan S. Turner 1998
"... In PAGE 5: ... Table 1 gives the parameter values of the fast forward counter at the beginning of a pass. Table2 shows the selection process. Table 1: Parameters of the Fast Forward Counter Current Counter Previous Counter Changing Bit Mask CarryIn 0100 0011 0111 1011... ..."
Cited by 6

Table 3.1 Correlation coefficients between line flows and MVAr margin

in for Transfer Limits
by Final Project Report, Anjan Bose, Robert Stuart, Ben Williams, Mark Willis, Liqiang Chen, Mohammad Vaziri

Table 5: The number of binary features and average time for each feature type. Feature types were broken into 23 tests for feature selection.

in
by Mark Dredze

Table 3: Selected di erences between Fast Messages and Myrinet API

in High Performance Messaging on Workstations: Illinois Fast Messages (FM) for Myrinet
by Scott Pakin, Mario Lauria, Andrew Chien
"... In PAGE 12: ...0, available in March 1995) and used to support their TCP/IP implementations. Table3 highlights some of the di erences between the two messaging layers. From the preceding discussion, it should be clear that adding even the smallest feature to the LCP can exact a large penalty in performance.... ..."

Table 3: Selected di erences between Fast Messages and Myrinet API

in High Performance Messaging on Workstations: Illinois Fast Messages (FM) for Myrinet
by Scott Pakin, Mario Lauria, Andrew Chien
"... In PAGE 12: ...0, available in March 1995) and used to support their TCP/IP implementations. Table3 highlights some of the di erences between the two messaging layers. From the preceding discussion, it should be clear that adding even the smallest feature to the LCP can exact a large penalty in performance.... ..."

Table 3. Classification results for our fast ratio feature selection algorithm using the SAM distance matrix in the NNB classifier.

in Hyperspectral Ratio Feature Selection: Agricultural Product Inspection Example
by Songyot Nakariyakul, David Casasent
"... In PAGE 7: ... We attribute this to differences between the training and test sets and to the small size of the bad training set (54 bad nuts vs 172 good nuts). We list in Table3 the same scoring measures and data using the SAM metric (rather than the EMD one) for our NNB classifier for the training and test set data. The SAM metric cannot be used with only one ratio feature in the NNB classifier; thus, we do not consider it for this case.... In PAGE 7: ... As before, generalization was good. From Table3 , when we chose two best features from the 50 FS ordered set, we note that one of the ratio features is 17/18; thus, adjacent bands provide useful information for classification. The best PC was again obtained when one ratio feature was chosen from the 50 FS ordered set and the other from the remaining 4900 ratio features.... ..."

Table 2. Classification results for our fast ratio feature selection algorithm using the EMD distance matrix in the NNB classifier.

in Hyperspectral Ratio Feature Selection: Agricultural Product Inspection Example
by Songyot Nakariyakul, David Casasent
"... In PAGE 7: ... We note that this single best ratio feature (using PC) was not in the top 50 ratio features (ordered by the criteria function J and the FS algorithm). As we see from Table2 , the PC scores for the training and test set using one or two ratio features using the EMD metric are comparable; thus, generalization is good. Using two ratio features yields significantly higher PC scores than using only one ratio feature as we expected.... In PAGE 7: ... The best PC was again obtained when one ratio feature was chosen from the 50 FS ordered set and the other from the remaining 4900 ratio features. Using the EMD distance metric ( Table2 ) in our NNB classifier gives about a 2% higher PC training score than does use of the SAM metric. The multiplicative scaling advantage of the SAM metric is not of use here, since our HS database was normalized and well recorded.... ..."

Table 2: Selection of the parameter K, using Local BinaryPattern feature space as a reference.

in Supervised texture classification for . . . tissue characterization
by Oriol Pujol, Petia Radeva
"... In PAGE 44: ... the possible geometryof samples in the feature space. Table2 illustrates the results regarding the selection of the number of neigh- bors k. It can be seen that for k = 7 a lower pixel error rate is obtained.... ..."

Table 5-2 Binary Scheduling Wheel Selection Process

in ABSTRACT DYNAMIC QUEUE MANAGEMENT FOR ATM NETWORK QUALITY-OF-SERVICE
by Yuhua Chen, Saint Louis Missouri, Yuhua Chen, Advisor Professor, Jonathan S. Turner, Yuhua Chen 1998
"... In PAGE 87: ...The following example shows how the algorithm works. Table5 -1 gives the parameter values of the fast forward counter at the beginning of a pass. Table 5-2 shows the selection process.... In PAGE 87: ... Table 5-1 gives the parameter values of the fast forward counter at the beginning of a pass. Table5 -2 shows the selection process. The value of the mask register Mask is (1011)2, which means only Wheel 2 is empty.... In PAGE 87: ... This process continues until all selected scheduling wheels have been served once. Table5 -1 Parameters of the Fast Forward Counter Current Counter Previous Counter Changing Bit Mask CarryIn 0100 0011 0111 1011 0001 Table 5-2 Binary Scheduling Wheel Selection Process... In PAGE 94: ... A credit table can be set up based on the link configurations. Table5 -3 shows how to set up the... In PAGE 96: ... When a virtual port is selected, one cell is allowed to forward to the link associated with it. In this way, the OC-48 link has eight chances to win an extra cycle as Table5 -3 Credit Table Link Number Link Type Basic Credits Extra Credits 0 OC-3 1 0 1 OC-3 1 0 2 OC-3 1 0 3 OC-3 1 0 4 OC-12 1 3 5 OC-12 1 3 6 OC-12 1 3 7 OC-12 1 3 8 OC-48 8 8 9-00 10 - 0 0 11 - 0 0 12 - 0 0 13 - 0 0 14 - 0 0 15 - 0... In PAGE 97: ... Figure 5-5 and Figure 5-6 shows the token passing circuitry. Table5 -4 Modified Credit Table Virtual Port Mask Actual Link Number Link Type Basic Credits Extra Credits Non-Empty 0000 1111 0 OC-3 1 0 1 0001 1111 1 OC-3 1 0 1 0010 1111 2 OC-3 1 0 1 0011 1111 3 OC-3 1 0 1 0100 1111 4 OC-12 1 3 1 0101 1111 5 OC-12 1 3 1 0110 1111 6 OC-12 1 3 1 0111 1111 7 OC-12 1 3 1 1000 1000 8 OC-48 1 1 1 1001 1000 8 - 1 1 1 1010 1000 8 - 1 1 1 1011 1000 8 - 1 1 1 1100 1000 8 - 1 1 1 1101 1000 8 - 1 1 1 1110 1000 8 - 1 1 1 1111 1000 8 - 1 1... ..."
Results 1 - 10 of 54,552