### TABLE II OUTLIER DETECTION PERCENTAGE IN RANDOM DESIGN.

### Table 2: Synopsis of results { quality of approximation

1998

Cited by 125

### Table 2. Interesting Outliers, Discriminants and the Performance of OBE. OO denotes outstanding outliers, IO denotes interesting outliers. Precision(Preci-), recall(Reca-) and the number of iterations(Iter-) for convergence in the classi cation step are used to show the performance of OBE.

2004

"... In PAGE 9: ... OBE reports as interesting outliers the outstanding ones, as well as those returned by the classi er. Table2 shows all the sets of interesting outliers along with the corresponding discriminants used as the underlying outlier concept in our experiments. In the table, for instance, the discriminant ( 1, 35, gt;, 0.... In PAGE 9: ...9 in the range of radii from 1 to 35. The number of the outstanding outliers and interesting outliers is also shown in Table2 . We always randomly sample 10% (y = 10) of the interesting outliers to serve as user- provided examples and hide the rest.... In PAGE 10: ... On the top row, we show the interesting outliers, original examples and the detected results for case U-Fringe. The bottom row shows those for case U-Corner (see Table2 for a description of the cases). Note that the chosen features can capture the notion of both edge and corner and, furthermore, OBE can almost perfectly reconstruct these hidden outlier notions! Ellipse dataset.... In PAGE 11: ...ig. 7. Detection Results on the Uniform Dataset. Top row: case U-Fringe, bottom row: case U-Corner see Table2 for description of each case. Fig.... In PAGE 11: ...ig. 8. Detection Results on the Ellipse dataset. From top to bottom, in turn: case E-Fringe, case E-Long, case E-Short see Table2 for description of each case. NYWomen dataset.... In PAGE 12: ...ig. 9. Detection Results on the NYWomen Dataset. From top to bottom in turn: Case N-FS, Case N-PF, Case N-SS see Table2 for description of each case. Only the rst and forth dimensions are used for the plots, although NYWomen Dataset is four dimensional.... In PAGE 12: ... Because of space limits, we only show the result plots in the rst and forth dimensions see Figure 9. For all datasets, Table2 shows the precision and recall measurements for OBE, using polynomial kernels (as mentioned, polynomial kernels always per- formed better than linear kernels in our experiments). It also shows the number of iterations needed to converge in the learning step.... In PAGE 12: ... It also shows the number of iterations needed to converge in the learning step. In Table2 , all the mea- surements are averages of ten trials. In almost all cases, OBE detects interesting outliers with both precision and recall reaching 80 90%.... ..."

Cited by 3

### Table 1 Detection rate for the 3 outliers located at time instants 127; 128; 129 for the OAD

2005

Cited by 6

### Table 2 Detection rate for the 3 outliers located at time instants 127; 128; 129 for the ROAD

2005

Cited by 6

### Table 2 Outlier detection performances on the UCI datasetsa

1999

"... In PAGE 6: ... To estimate the errors (of the first and the second kind) n-fold cross-validation with n 5 is used. In Table2 , the performances of the outlier de- tection methods on all UCI datasets are shown. For each method, the performance on a target validation set (left) and an outlier test set (right) is shown.... ..."

Cited by 59

### Table 1.1 Number of iterations needed to solve the second linear system when the reuse of vectors in the subspace is done by truncation, assembling (rank-k) or assembling of preconditioned Ritz vectors (k outliers) for the rectangular basin partitioned in 4 strips.

### Table 1 Number of iterations needed to solve the second linear system when the reuse of vectors in the subspace is done by truncation, assembling (rank-k) or assembling of preconditioned Ritz vectors (k outliers) for the rectangular basin partitioned in 4 strips.

"... In PAGE 7: ...ossible. Note that the construction is also based on (4.13). Based on the results in Table1 the assembling strategy of preconditioned Ritz vectors has been chosen for further experiments. 6 Test Case and Results The test case is concerned with the ow in a 8000m by 1200m rectangular basin which is 8m deep.... ..."

### Table 1: Arti cial time series data without outliers.

"... In PAGE 3: ... Consider the outlier-contaminated time series shown in Figure 1. The outlier-free data consist of a random realization of n = 50 observations given in Table1 and generated from the AR(3) model, xt = 8 gt; lt; gt; : at t = 1, 2, 3 2:1xt?1 ? 1:46xt?2 + 0:336xt?3 + at t = 4; : : : ; 50; where fatg is a sequence of independent and identically distributed Gaussian variates with mean zero and variance 2 a = 1: The roots of the autoregressive polynomial are 0.... ..."

### Table 6 lists the processing results a fter applying the proposed multiple outlier detection procedure

"... In PAGE 15: ...0 %) 350 ( 4 . 6 %) Table6 . Single-epoch solution using the multiple outlier detection strategy and the real-time stochastic model derived using residuals from previous epochs.... ..."