Results 1 - 10 of 20,353

Table 2: A summary of the differences between fixed size partitioning and tag-based partitioning.

in Efficient Update of Indexes for Dynamically Changing Web Documents
by Min Wang, Jeffrey Scott Vitter, Lipyeow Lim, Sriram Padmanabhan, Ramesh Agarwal

Table 2: Edit word detection performance for two word-based PCFGs and the TAG-based edit detec- tor. F -score and error are word-based measures. Edit Detector Edit F -score Edit Error

in Effective Use of Prosody in . . .
by Jeremy G. Kahn, Matthew Lease, Eugene Charniak, Mark Johnson, Mari Ostendorf
"... In PAGE 6: ...5 The TAG-based de- tector was trained on the same conversation sides, with its channel model trained on the Penn Treebank disfluency-annotated files and its language model trained on trees with the EDITED nodes excised. As shown in Table2 , we did find that both PCFGs per- formed significantly below the TAG-based detector. 5 Results In evaluating parse accuracy, we adopt the relaxed edited revision (Charniak and Johnson, 2001) to the standard PARSEVAL metric, which penalizes sys- tems that get EDITED spans wrong, but does not pe- nalize disagreements in the attachment or internal structure of edit regions.... ..."

Table 1: Query results for the three types of query formulation Query

in Video Data Modelling To Support Hybrid Query
by Lilly Suriani Affendey, Ali Mamat, Hamidah Ibrahim, Fatimah Ahmad
"... In PAGE 8: ... Based on these three comparisons, we summarize the results for the three query types. Table1 summarizes the query results for the three types of query formulation based on colour weight 0.... ..."

Table 1. A comparison between three optical switching paradigms OBS protocols can be roughly classi ed into two types: JET-based and TAG-based [9,10]. The former uses an o set time, T, between each burst and its control packet, but the latter does not. Speci cally, using JET, a source sends out a control packet, which is followed by a burst after T PH h=1 (h), where (h) is the (expected) control delay (i.e., the processing time incurred by the control packet) at hop 1 h H. Because the burst is bu ered at the source (in the electronic domain), no FDLs are necessary at each intermediate node to delay the burst while the control packet is being processed (but such FDLs are necessary when using TAG-based protocols). In addition, JET uses delayed reservation (DR) to e ciently utilize the bandwidth. For example, as shown in Figure 1, the bandwidth on hop i is reserved from the time the burst is expected to arrive, i.e., tb = tc + T(i), where tc is the time the control packet arrives and T(i) = T ? P(i?1) h=1 (h).

in A New Optical Burst Switching Protocol for Supporting Quality of Service
by Myungsik Yoo, Chunming Qiao
"... In PAGE 3: ...Table1... ..."

Table 1 lists the kl divergence between the true and learnt model, as well as the number of runs until convergence was reached, for each of the 5 sequences for both the setting that uses odometric information under tag-based initialization and the learning algorithm that does not use odometric information, averaged over 10 runs per sequence. We stress that each kl divergence measure is calculated based on new data sequences that are generated from the true model, as described in Section 6.2. The 5 sequences from which the models were learnt do not participate in the testing process. The kl divergence with respect to the true model for models learnt using odometry, is about 5-6 times smaller than for models learnt without odometric data. The standard deviation around the means is about 0.2 for kl distances for models learnt with odometry and 1.5 for the no- odometry setting. To check the signi cance of our results we used the simple two-sample t-test. The models learnt using odometric information have statistically signi cantly (p 0:0005) lower average kl divergence than the others.

in Learning Geometrically-Constrained Hidden Markov Models for Robot Navigation: Bridging the Topological-Geometrical Gap
by Hagit Shatkay, Leslie Pack Kaelbling 2002
"... In PAGE 31: ... Table1 : Average results of two learning settings with ve training sequences.... ..."
Cited by 4

Table 8: The performances of image retrieval

in Cross-Language Image Retrieval via Spoken Query
by Wen-cheng Lin, Ming-shun Lin, Hsin-hsi Chen 2004
"... In PAGE 10: ... The selected English terms were submitted to Okapi IR system to retrieve English captions. The retrieval performances are shown in Table8 . In order to compare the performances of spoken query with textual query, we conduct two runs that using original Chinese textual queries.... In PAGE 10: ... The performance of spoken query is about 45% and 40% of textual query in CO model and F2HF model, respectively. From Table8 and 9, F2HF model performs better than CO model. In F2FH model, the top two most frequent translations are selected.... ..."
Cited by 4

Table 16: Medical image retrieval systems

in fMRI neuroinformatics
by Finn ˚arup Nielsen, Acknowledged Are Lael Gatewood 2006
"... In PAGE 33: ..., 2006). Table16 lists other content-based image retrieval systems in the neuroimaging context. Table 16: Medical image retrieval systems... ..."
Cited by 1

Table 16: Medical image retrieval systems

in fMRI neuroinformatics
by Finn ˚arup Nielsen, Acknowledged Are Lael Gatewood 2006
"... In PAGE 33: ..., 2006). Table16 lists other content-based image retrieval systems in the neuroimaging context. Table 16: Medical image retrieval systems... ..."
Cited by 1

Table 1: Number of retrieved images for different queries and the % of relevant images in the database. Number of retrieved images in setup

in Modeling User Context with Applications to Media Retrieval
by Ankur Mani, Hari Sundaram 2007
Cited by 2

Table 2 Image features used for retrieval

in Locality pursuit embedding
by Wanli Min A, Kelu B, Xiaofei He C 2003
"... In PAGE 7: ... A retrieved image is considered correct if it belongs to the same category of the query image. Three types of color features and three types of texture features are used in our system, which are listed in Table2 . Each image is represented by a 435-dimensional vector in the image space.... ..."
Results 1 - 10 of 20,353