### Table 1 VP and RNN learning curves in the first pass attachment predictio POS Noun Verb Prep. Article Punc. Adjective

2005

"... In PAGE 8: ... Fixing the maximum outdegree is tions are typically needed to complete the training procedure. The results of the comparison are shown in Table1 , where the error measure is based on the L1 loss of Eq. (8) which counts the number of for- ests where the best element is not ranked first.... ..."

### Table 4 The table is in agreement with the fact that each curve has a three-fold point and passes

### Table 1: Arguments passed into and Values Returned by a User-Supplied Subroutine for the Parametrization of Curves and Surfaces

"... In PAGE 5: ... CMPGRD passes in to this subroutine the appropriate input parameters scaled to the unit interval (for two-dimensional curves) or unit square (for three-dimensional surfaces), and in addition, passes in the cosine and sine of the angular parameters and/or for coordinate systems where these are applicable. CMPGRD expects the subroutine to return the value of the function c and its derivative in a form appropriate to the coordinate system, as detailed in Table1 .... ..."

### Table 4 shows the extreme and singular points that we passed while a curve is traced. Note that the tangential point does not affect the marching behavior.

"... In PAGE 7: ... Table4 : Representative Points of CBBGBD A0 CBBGBE BP BC. The fifth pair of surfaces CBBHBDB4D9BN DAB5=B4D9BN DABN B4BDBPBIB5 A3 D9BI A0 B4BHBPBGB5 A3 D9BG B7 BE A3 D9BE B7 DABEB5 CBBHBEB4D7BN DBB5=B4D7BN DBBN BCBMBLBDBJB5 results in a 2-eight-shaped curve (Figure 15).... ..."

### Table 4 shows the extreme and singular points that we passed while a curve is traced. Note that the tangential point does not affect the marching behavior.

"... In PAGE 7: ... Table4 : Representative Points of CB BGBD A0 CB BGBE BPBC. The fifth pair of surfaces CB BHBD B4D9BN DAB5=B4D9BN DABN B4BDBPBIB5 A3 D9 BI A0 B4BHBPBGB5 A3 D9 BG B7BEA3 D9 BE B7 DA BE B5 CB BHBE B4D7BN DBB5=B4D7BN DBBN BCBMBLBDBJB5 results in a 2-eight-shaped curve (Figure 15).... ..."

### Table 1. Classification of frequency-tuning curves. Maximum and vector strength (VS) tuning curves of all three network types were classified as untuned, low pass, high pass, trough, peaked or complex

2000

"... In PAGE 6: ... 8f). All six classes could be found in the tuning curves ( Table1 ). While the high pass and peaked class were rare, the other four classes were quite abundant in all three net types.... ..."

### Table 5. Bifurcation curve labels for gures 2{7 To complete our work, we consider the curve of Hopf bifurcation of the mixed mode, given by (17)-(18). We begin by noting that this complicated curve passes through the origin tangent to the line 1 = ? 1=a2 22 0 2 = a11(a12 ? a22)=[a22(a11 ? a21)] 2 and asymptotically behaves as the parabola 1 ?a122 a2 22 2 2. In principle, conditions similar to those of (22), (24), (26) may be formulated for the Hopf bifurcation by requiring that at least one of the mixed modes exist and the determinant of the linearization at this mixed mode (16) be negative. Deriving these conditions is quite involved, thus

### Table 4: Generated candidates. While the area beneath the curve corresponds to the total volumes, the height of the curve at each point describes how much processing is needed during that pass. High peaks correspond to passes where the overlapping of I/O and processing may be endangered if the database is large. Since the con dence threshold only a ects the number of rules generated from the covering sets we have not varied it in our experiments. On the other hand, suitable values for the support threshold depend very much on the

1994

"... In PAGE 12: ... The number of passes for each algorithm is presented in Table 3: within this data OCDs sometimes requires less passes than the others, but the number of passes obviously depends strongly on the data that is studied. Table4 presents the number of candidates considered by each method. The numbers for AIS are much higher, as it may generate the same candidates over and over again during the database pass.... ..."

Cited by 5

### Table 2. Activities of PASS 1.41 That Were Defined as Being Associated with Antineoplastic Action

2002

"... In PAGE 7: ...and this entity, which can be seen as a measure of the preponderance of a database for a given subset of activities, is plotted, for varying cutoff points Pi cutoff of the Pi range admitted, in Figures 6 and 7. These curves were calculated for both the subset of 56 activities that can be associated with antineoplastic action (Figure 6), which are listed in Table2 , and for the remaining 509 of the total 565 PASS v. 1.... In PAGE 7: ... 1.41 activities ( Table2 ). Squares: NCI Open DB (upper curve); diamonds: ACD (lower curve).... ..."

### Table 1: Algorithm for generating an ROC curve from a set of ranked examples.

2001

"... In PAGE 6: .... For each classifier, plot TP and FP in ROC space. For continuous-output classifiers, vary a threshold over the output range and plot the ROC curve. Table1 shows an algorithm for producing such an ROC curve in a single pass.3 2The convex hull of a set of points is the smallest convex set that contains the points.... In PAGE 6: ...3 2The convex hull of a set of points is the smallest convex set that contains the points. 3There is a subtle complication to producing ROC curves from ranked test-set data, which is reflected in the algorithm shown in Table1 . Specifically, consecutive examples with the same score can give overly optimistic or overly pessimistic ROC curves, depending on the ordering of positive and negative examples.... In PAGE 16: ... The rocch method can be used to organize such ranking models, as we have seen. Recall that ROC curves are formed from case rankings by moving the cutoff from one extreme to the other ( Table1 shows an algorithm for calculating the ROC curve from such rankings). The rocch-hybrid comprises the ranking models that are best for all possible conditions.... ..."

Cited by 170