Results 1 - 10 of 7,604

Table 3. Parameter values em- ployed in the experiments with

in Applying Genetic and Symbolic Learning Algorithms to Extract Rules from Artificial Neural Networks
by Claudia Regina Milaré, Gustavo E. A. P. A. Batista, André C. P. L. F. Carvalho, Maria Carolina Monard 2004
"... In PAGE 5: ... Phase 2 was performed as follows: 1. TREPAN was executed with different val- ues assigned to its main parameters, in- cluding its default parameters, as showed in Table3 . The MinSample parameter of TREPAN specifies the minimum num- ber of instances (i.... ..."
Cited by 1

Table 3. Parameter values em- ployed in the experiments with

in Applying genetic and symbolic learning algorithms to extract rules from artificial neural neworks
by Claudia R. Milaré, Maria C. Monard 2004
"... In PAGE 5: ... Phase 2 was performed as follows: 1. TREPAN was executed with different val- ues assigned to its main parameters, in- cluding its default parameters, as showed in Table3 . The MinSample parameter of TREPAN specifies the minimum num- ber of instances (i.... ..."
Cited by 1

Table 3: The 10 best consensus features of the NN classifier as a function of the time stack lag, L. The DPCA transform was em- ployed.

in Decision time horizon for music genre classification using short time features
by Peter Ahrendt, Anders Meng, Jan Larsen 2004
Cited by 9

Table 1: Maximum Errors for the approximation of a Poisson equation on an in nite domain. For each approximation 16 subdomains are em- ployed, and the polynomial degree is given for each trial. For each trial the polynomial degree for both the x and y direction are equal.

in Spectral Element Approximations and Infinite Domains
by Kelly Black
"... In PAGE 14: ... Away from the unit square semi-in nite subdomains are employed to construct an approximation (see Figure 11). The errors are shown in Table1 , and the errors are the max- imum errors found on the abscissa of the Legendre-Gauss quadrature on each subdomain.... ..."

Table 2: E ect of di erent speeds of components results based on a complex component library. We em- ployed MACs which are able to perform the addition and multiplication (separately or as macro-operation) within one control step.

in OSCAR: Optimum Simultaneous Scheduling, Allocation and Resource Binding Based on Integer Programming
by Birger Landwehr, Birger L, Peter Marwedel, Rainer Dömer
"... In PAGE 5: ... In the following, we present experimental results for the 5th-order Elliptical Wave Filter [9] and the Di erential Equation Solver benchmark. 5th-Order Elliptical Wave Filter Table2 shows the results of the EWF employing ad- ders, multipliers and multi-functional units with di e- rent delays, latencies and costs. All delays are measu- red in control steps.... ..."

Table 4.5: Correlation coefficients (CC [%]) and relative errors (RE [%]) of results em- ploying the FEM versus the analytical solution for the spherical model assuming electrical anisotropic conductivities within the source region. DL indicates the different discretiza- tion levels.

in Acknowledgement
by Dipl. -ing Michael Seger

Table 4 displays system performance using unigram fea- tures separately and combined with bigram features. When applicable our approach is compared to that of Zissman [8] (who uses ergodic HMMs) and Muthusamy [4] (broad- category segmentation) For trilingual LID a di erent ar- chitecture combining three bilingual experts was also em- ployed in order to let the architecture re ect the structure of the problem (Fig. 2). Performance improved slightly from 73.2% to 74.5% (unigrams only).

in unknown title
by unknown authors 1994
"... In PAGE 4: ...2% 74.2% - - Table4 . Results 5.... ..."
Cited by 11

Table 1 presents the average values of these control variables separately for em- ployed wives and housewives as well as t-tests assessing the statistical significance of differences between these two groups. The two groups of women differ significantly on education, family income, and number of children at home. On average, homemakers report less education, lower family incomes and more children living at home. Homemak- ers and employed wives do not differ

in Women, Work, and Well-Being: The Importance of Work Conditions*
by Mary Clare Lennon
"... In PAGE 6: ... dimensions examined. These results are M~~~~~~~, there is far less variability in largely maintained when controls are intro- reports of autonomy by homemakers; the duced for the social and demographic vari- standard deviation for homemakers is less ables listed in Table1 .5 A next step calls for than half that of employed wives on this evaluating the consequences of these differ- measure.... ..."

Table 1: A comparison of Symmetric-Galerkin and collocation nodal errors for a Dirichlet problem on the unit disk. For the Dirichlet boundary conditions, the Symmetric-Galerkin method em- ploys the potential equation. To examine the performance of the hypersingular ux equation, a Neumann problem is solved exterior to the unit disk. The cho- sen exact solution is = x=(x2 + y2), which also turns out to be the form of the normal ux boundary condition. The errors in the computed potential are shown in Table 2, and once again Symmetric-Galerkin is more accurate than collocation. Note that the collocation solution was obtained using the potential equation { the hypersingular equation cannot be used with the C0 linear element approximation.M Symmetric-Galerkin Collocation Max

in Evaluation of singular and hypersingular Galerkin integrals: direct limits and symbolic computation
by L. J. Gray 1998
"... In PAGE 16: ...Table1 compares the maximum and L2 errors in the computed normal ux for Symmetric-Galerkin and collocation approximations, both employing a linear element approximation. The L2 error is de ned as 1 M M X k=1 e2 k!1=2 ; (27) where M is the number of nodes and ek is the error at node k.... ..."
Cited by 6

Table 2: Average, maximum and standard deviation of E errors in predicting the test chart (with n and optimized) after dot growth function estimation and further primary correction (with multi-colorant samples 6 6 6) - (1) LS (LS G FC) and TLS (TLS G FC) regression for the rotated screen printer, em- ploying single-colorant and gray step-wedges for the dot growth function estimation ; (2) LS (LS GM FC) and TLS (TLS GM FC) for the rotated screen printer, employing single-colorant, gray and multi-colorant step-wedges for the dot growth function estimation; (3) LS (LS GD FC) and TLS (TLS GD FC) for the dot-on-dot screen printer, employing single-colorant and gray step-wedges for the dot growth function estimation.

in End-to-End Color Printer Calibration by Total Least Squares Regression
by Minghui Xia, Eli Saber, Gaurav Sharma, A. Murat Tekalp 1999
"... In PAGE 20: ... It is also possible to perform further primary estimation after employing single-colorant, gray and/or multi-colorant step-wedges to estimate the dot area functions. These results are shown in Table2 with the optimum YN correction factor n and the noise factor (for the combined dot-on-dot model). Again, the TLS based results produce smaller E errors than their LS counterparts.... In PAGE 20: ... Again, the TLS based results produce smaller E errors than their LS counterparts. Comparing Table2 with Table 1, where TLS primary correction was performed on only four primaries, it is clear that further primary correction improves the accuracy of the Neugebauer model predictions. 5 Conclusions This paper addresses the use of spectral Neugebauer models to characterize color halftone printers.... ..."
Cited by 4
Results 1 - 10 of 7,604