### Table 5: Performance when provided only with non-identical fragments, with all other seg- ments identi ed.

1994

"... In PAGE 18: ... In such a case, the approximate length of the fragments are known, but not the multiplicity. For the experiments reported in Table5 , all fragments clustered together within the given grouping percentage were replaced by one instance of the mean length of these fragments. Thus, unlike our other tables, the problems get harder with increasing grouping percentage.... In PAGE 18: ... Thus, unlike our other tables, the problems get harder with increasing grouping percentage. The number of missing fragments is reported within parentheses in Table5 . All missing fragments were assumed to be multiple occurrences of existing fragments instead of unrestricted wildcards.... In PAGE 18: ... Five random instances of N base fragments were generated and subjected to a random relative error of r, where 10 N 20 and 0:0% r 2:5%. Table5 summarizes these experiments, by presenting the con dence interval with = 0:05 for the average running time in seconds. It clearly illustrates that the critical region is r = (1=n2), as is predicted by the theory.... ..."

Cited by 12

### Table 4: Number of tuples and non-identical dependency triples (types) extracted per de- pendency relation.

Cited by 1

### Table 3.1: Optimization results using identical and non-identical thresholds

2004

### Table 2 Comparison of Scenarios for Non-identical Firms with Different Cross Sensitivities.

2001

"... In PAGE 18: ... On the other hand, if the net effect is positive, Firm 2 generates higher profits under DC than CC and under DD than CD. We can observe this ordering of scenarios for Firm 2 in the example in Table2 . As the percentage of customers Firm 1 loses through lead-time competition is high, Firm 2 benefits from the longer lead-times quoted by a decentralized competitor and generates the highest profits under DC.... ..."

### Table 5: Sample Queries for Non-Identical Dups natural language queries, often due to the perceived con- trol it offers users. The default results ranking for Boolean queries on Westnews is by date (i.e., reverse chronologi- cal order). This characteristic permits an initial binning- by-date that can be exploited later when avoiding costly term vector comparisons. A similar economy can be estab- lished by binning-by-doc length within each date-based bin.

2003

Cited by 15

### Table 1: Known results for batch scheduling of con icting jobs under minsum criteria

2002

"... In PAGE 4: ... Our results for MSJCT yield signi cant improvements over previous bounds; in particular, the bounds that we derive for interval graphs, line graphs and perfect graphs improve upon the uniform bound of 16 obtained for these graph classes in [2]. We summarize the known results for MSJCT in Table1 . New bounds given in this paper are shown in boldface, with the previous best known bound given in parenthesis.... ..."

Cited by 2

### Table 5: Results for real maps, without ltering out non-identical matches. The RMS calculation pairs up neighboring atoms and takes the average distance be- tween such pairs.

"... In PAGE 7: ... For example, there might be a good match of a Gln to an Asp with high density correlation that would be rejected in favor of the best Asp in the database, which might have a much lower correlation. Therefore, in Table5 we show results for constructing models for the real maps of crambin and avodoxin as we intended for TEXTAL. The models constructed now do not have the same amino acid sequences as the original structures.... ..."

### Table 3: Error of the fixed time window method with fixed window sizes from BDBMBMBMBH batches for splits of the business cycle data into 5 and 15 batches, respectively.

"... In PAGE 7: ... Only the fixed size time window approach can the- oretically compete, if the optimal fixed time window size is known in advance, which is usually not possible in a real- world application. As shown in Table3 , using other fixed window sizes, leads to significant drops in performance to error levels, mostly well above those of the two adaptive approaches. The fact that the fixed size approach is competitive in this domain may be due to the cyclic nature of the domain.... ..."

### Table 3: Error of the fixed time window method with fixed window sizes from 1:::5 batches for splits of the business cycle data into 5 and 15 batches, respectively.

in Meta-Learning, Model Selection, and Example Selection in Machine Learning Domains with Concept Drift

"... In PAGE 7: ...known in advance, which is usually not possible in a real- world application. As shown in Table3 , using other fixed window sizes, leads to significant drops in performance to error levels, mostly well above those of the two adaptive approaches. The fact that the fixed size approach is competitive in this domain may be due to the cyclic nature of the domain.... ..."

### Table 2: Measured throughput and efficiency for various settings of concurrency bound and batch size.

2007

"... In PAGE 9: ... Our goal is to show that we can adjust these parameters to obtain high I/O effi- ciency. Table2 shows the measured throughput and efficiency met- rics for different parameter values, of workload mixes RRR, RLL, LLL and SSS respectively. These results show that the baseline DRR scheduler (where D = 1 and G = 1) does indeed exhibit poor I/O efficiency, between 0.... ..."