Tools
Sorted by:
Try your query at:
 |
 |
 |
 |
 |
 |
Results 1 - 10
of
16,151
Table 1: Average retrieval response times (in seconds) for sequential search of the entire database.
2002
"... In PAGE 17: ...1 Efficiency The purpose of this set of experiments is to (a) Identify the faster method and (b) Identify the faster variant from each category of methods. Table1 shows the average retrieval response times obtained by all methods grouped by category. These are times for sequential search of the entire database.... ..."
Cited by 13
Table 3: Average results over all hold-out samples using the entire database
1999
"... In PAGE 8: ...8 Comparison of Modeling Techniques Table3 and Table 4 show the cross-validation results of comparing the modeling techniques in terms of MMRE, MdMRE, and Pred(.25) for the whole data set and for the bank data, respectively.... In PAGE 8: ... The values in the tables are averages across the hold-out samples. In Table3 the hold- out samples (test sets) are the six companies with more than 10 projects. In Table 4 the hold-out samples are randomly selected samples.... In PAGE 8: ... CART+Analogy-1s and CART+Analogy-2s are a combination of CART with Analogy-1s and Analogy-2s, respectively. Considering the whole database ( Table3 ), we observe that, on average, the techniques not involving analogy (i.e.... In PAGE 12: ... Table A-2 summarizes the results based on the bank data (6 random samples). The averages (last three rows from each table) are identical to Table3... ..."
Cited by 23
Table 3: Classi cation accuracies (in %) with and without incremental learning; Case i represents one of the incremental learning methods and Non- Incremental represents the case where the entire database was used to train the classi er.
1999
"... In PAGE 4: ... The performance of the classi er was then compared to that of a classi- er trained on the entire set of database images (non- incremental learning). Table3 shows the classi cation accuracies for the various classi ers with and without incremental learning. The best classi cation accura- cies achieved for each of the classi ers were 95:9% for the city vs.... ..."
Cited by 1
Table 1. Misclassification rates (%) for motion database Entire dataset with 2 and 3 motions LLMC LLMC MSL GPCA LSA LSA CCS
2007
"... In PAGE 5: ... We refer to these variants as LLMC5, LLMC4m,LSA5andLSA4m. Table1 contains the average and median classification errors given by each algorithm on the 155 sequences, and Figure 1 shows histograms of these errors. Notice that MSL gives nearly perfect segmentation for the majority of the se- quences, but it occasionally gives large errors when it con- verges to a local minimum.... ..."
Cited by 1
Table 25.3: Precision at the global and local-level for the entire image database; the overall performance is the weighted average of the retrieval performance by organ (each image was a query image)
Table 2.1 shows the expected result of using the learned recognition model on the entire SWISS-PROT database. From Equation 3 and Table 2.1 it follows that:
2000
"... In PAGE 3: ...Table2 . 2 2 Contingency table for SWISS-PROT.... ..."
Cited by 3
Table 1b shows the expected result of using the learned recognition model on the entire SWISS-PROT database. From Equation 3 and Table 1b it follows that:
"... In PAGE 3: ... Let C = the cost of test- ing the biological activity of one protein via wet-experiments in the laboratory; NPP = Sequence is a NPP; Rec = Model recognises sequence as a NPP. RA = C=Pr(NPP) C=Pr(NPP j Rec) = Pr(NPP j Rec) Pr(NPP) (1) Let testing the model on test data yield the 2 2 contingency table shown in Table1 a with the cells n1, n2, n3, and n4. Let n = n1 + n2 + n3 + n4 be the number of instances in the test-set.... In PAGE 4: ...Table1 . 2 2 Contingency table for a) the test-set and b) SWISS-PROT.... ..."
TABLE I AVERAGE RUNTIME FOR EACH STAGE OF THE BLAST ALGORITHM FOR 100 RANDOMLY-SELECTED SEQUENCES FROM THE GENBANK NON-REDUNDANT DATABASE SEARCHED AGAINST THE ENTIRE DATABASE. EXPERIMENTS WERE CONDUCTED USING THE NCBI IMPLEMENTATION OF BLAST AND DEFAULT PARAMETERS, INCLUDING W = 3, T = 11, S1 = 22 BITS, UNGAPPED DROPOFF OF 7 BITS, GAPPED DROPOFF OF 15 BITS AND BLOSUM62 SUBSTITUTION MATRIX WITH GAP PENALTIES OF 11 AND 1.
Table 2 shows the performance results when applied to the entire test database. It can be seen that the estimates are biased and the bias increases with d. A disadvantage of this method is the requirement of long data records. Further, the value of m from which (26) holds is arbitrary.
1998
"... In PAGE 20: ...d 0:1 0:2 0:3 0:4 mean( ^ d) 0:1068 0:1857 0:2637 0:3351 var( ^ d) 0:0018 0:0021 0:0017 0:0012 R/S Method d 0:1 0:2 0:3 0:4 mean( ^ d) 0:1821 0:2369 0:2863 0:3315 var( ^ d) 0:0004 0:0004 0:0006 0:0004 Table2 : Estimation results using Aggregated Variance and R/S method of d was obtained as ^ d = 0:2472 while Figure 10(b) shows the pox plot for fARIMA(1,0:4,1) series and the estimate of d was obtained as ^ d = 0:3160. The value of lag k from which (30) holds seems to be arbitrary.... In PAGE 20: ... The value of lag k from which (30) holds seems to be arbitrary. Table2 shows the performance results obtained when applied to the test database. The estimates have a positive bias for small values of d and a negative bias for large values of d.... ..."
Cited by 1
Results 1 - 10
of
16,151
