### Table 1: Mutation Operators for Inter-Class Testing

2005

"... In PAGE 3: ... [43] developed a comprehensive set of class mutation operators for Java. This set of 24 mutation operators is summarized in Table1 . Each mutation operator is related to one of the following six language feature groups.... In PAGE 13: ... 3.3 The MuJava Tool The Java class mutation operators in Table1 have been implemented in a tool called MuJava (Mutation System for Java). MuJava is the result of a collaboration between two universities, Korea Advanced Institute of Science and Technology (KAIST) in South Korea and George Mason University in the USA.... ..."

Cited by 34

### Table 6 Inter-class distances between the differenced time series from the decreasing trend and the upward shift classes

2006

"... In PAGE 20: ... As a matter of fact, as we compute the pairwise distances between all 6 differ- enced series, we realize that the distances are not indicative at all of the classes these data belong. Table 5 and Table6 show the inter- and the intra-distances between the series (the series from the decreasing trend class are denoted as... ..."

### Table 1. Comparison of Performance in Student-Centered and Teacher-Centered Sections of Undergraduate Thermodynamics.

### Table 4: Inter-class variation vs. intra-class variation for point set B-1. Points and thermophysical properties are chosen for the car hypothesis. The middle column shows feature values computed when the measurement tensor is obtained from the van. This set exhibits poor inter-class separation.

"... In PAGE 19: ... Table 3 shows inter-class and intra-class variation when a van is hypothesized, and for images obtained at ve di erent times in the day. Table4 shows inter-class and intra-class variation when the car is hypothesized. Such investigation showed that the set of points B-1 produced almost identical values irrespective of the source of the measurements.... ..."

Cited by 1

### Table 3: Inter-class variation vs. intra-class variation for set A-1. Points and thermophysical prop- erties are chosen for a van hypothesis. The middle column shows feature values computed when the measurement tensor is obtained from the car.

"... In PAGE 19: ... coordinates of the van points (in the van center coordinate frame) to the image frame computed for the unknown vehicle. Table3 shows inter-class and intra-class variation when a van is hypothesized, and for images obtained at ve di erent times in the day. Table 4 shows inter-class and intra-class variation when the car is hypothesized.... ..."

Cited by 1

### Table 3. Inter-class recovery transition hit rate of programmers, k = 4.

"... In PAGE 8: ... While momentum offers an intuitive way of examining the importance of a method, it does not perform well in periods of low intensity or searching. In the experiment described for Table3 , the LRU method replacement algorithm was used to evaluate the recovery of previously visited methods for ten programmers A - J. Eval- uating recovery is subtly different than inter-class transition hit rate.... ..."

### Table 1: Inter-class separation, mean-square representation error and classi#0Ccation accuracy

in A General Methodology for Simultaneous Representation and Discrimination of Multiple Object Classes

1998

"... In PAGE 11: ...classi#0Ccation of the test set projections on the three basis vectors. The results are shown in Table1 . The MRDF has the best classi#0Ccation rate P C #28the precentage of test set data correctly classi#0Ced#29.... In PAGE 11: ... The MRDF has the best classi#0Ccation rate P C #28the precentage of test set data correctly classi#0Ced#29. The mean squared representation error and the mean squared separation in the test set projections for each of the three LDFs are listed in Table1 for completeness. As seen, the MRDF has a larger separation and a smaller representation error than the other LDFs.... In PAGE 11: ... The FK basis vector chosen was the one that best represented class 1 #28it thus chooses a vector such that the spread of the projections of class 1 is larger than the spread of the class 2 projections#29. As seen, the sample projections from the top clusters in classes 1 and 2 overlap; this results in its low P C #28 Table1 #29. The MRDF basis vector #28the dominant eigenvector in Eq.... In PAGE 18: ...Nonlinear MRDF Results We expect our nonlinear MRDFs to provide improved P C for the data in Fig. 1, but this has not been veri#0Ced since our linear MRDFs gave P C apos; 99#25 #28 Table1 #29. To test the nonlinear MRDF, we #0Crst consider two 2-D #28two features#29 synthetic data cases, in which the decision surfaces can be visualized.... ..."

Cited by 17

### Table 2. Inter-classes substitutions made by C5C4C8 on CBCTD8BE

"... In PAGE 3: ...e. the pairs of classes for which the addition of B4BVCXBN BVCYB5 and B4BVCYBN BVCXB5 MLP substitu- tions rates mentioned in Table2 reaches a fixed thresh- old rate). The set containing all this pairs is denoted CBC5CPCYCBD9CQC8CPCXD6D7 .... In PAGE 3: ...7,1), (9,3), (6,0), ...etc.) which constitutes the majority of the MLP substitutions, as shown by Table2 , were im- plemented by the software SVM-LIGHT provided by T. Joachims B4CWD8D8D4 BM BPBPCPCXD7BMCVD1CSBMCSCTBPAOD8CWD3D6D7D8CTD2BPB5.... ..."

### Table 1 Ratio of the averaged inter-class and intra-class distances between sample vectors: We show results for four different cut-off values with logged raw data, GM nor- malised data and LS normalised data.

"... In PAGE 11: ... Both datasets show a significant rotation and we can therefore conclude that the inclusion of many noisy, low expression readings will obscure the signal within the data. Figure 4 here Table1 shows the ratio of the mean Euclidean distance between expression vectors DCCX from different classes to the mean intra-class distance. We expect this value to be greater than one, indicating greater inter-class distances on average, although the results show that the effect is quite small.... In PAGE 11: ... It appears that the two normalisation methods are comparable, with LS normalisation providing slightly larger ratios than GM for dataset A and vice-versa for dataset B. Table1 here 3.3 Adaptive CP The normalisation model in equation (1) includes an additive term CQCX and a mul- tiplicative factor CPCX.... ..."

### Table 1: Inter-class separation, mean-square representation error and classi cation accuracy using MRDF, FK and Fisher discriminant functions.

in A General Methodology for Simultaneous Representation and Discrimination of Multiple Object Classes

1998

"... In PAGE 11: ...classi cation of the test set projections on the three basis vectors. The results are shown in Table1 . The MRDF has the best classi cation rate PC (the precentage of test set data correctly classi ed).... In PAGE 11: ... The MRDF has the best classi cation rate PC (the precentage of test set data correctly classi ed). The mean squared representation error and the mean squared separation in the test set projections for each of the three LDFs are listed in Table1 for completeness. As seen, the MRDF has a larger separation and a smaller representation error than the other LDFs.... In PAGE 11: ... The FK basis vector chosen was the one that best represented class 1 (it thus chooses a vector such that the spread of the projections of class 1 is larger than the spread of the class 2 projections). As seen, the sample projections from the top clusters in classes 1 and 2 overlap; this results in its low PC ( Table1 ). The MRDF basis vector (the dominant eigenvector in Eq.... In PAGE 18: ...Nonlinear MRDF Results We expect our nonlinear MRDFs to provide improved PC for the data in Fig. 1, but this has not been veri ed since our linear MRDFs gave PC apos; 99% ( Table1 ). To test the nonlinear MRDF, we rst consider two 2-D (two features) synthetic data cases, in which the decision surfaces can be visualized.... ..."

Cited by 17