Results 11 - 20 of 62,290

Table 4. Average BDeu scores. (D7 BP BG, mfss BP BG; BEBHBCBN BCBCBC random edges considered for hillclimbing) dataset rand hlclmb CBBUC6CB CBBUC6CB+MIe CBBUC6CB+MIe+2nd CBBUC6CB+MIe+2nd+hlclmb

in Tractable learning of large bayes net structures from sparse data
by Anna Goldenberg 2004
"... In PAGE 5: ...We tested our algorithm in a variety of configurations on the datasets listed in Table 3. The results in Table4 are reported in terms of the average BDeu score, i.e.... ..."
Cited by 19

Table 5. Number of links in the resulting nets. (D7 BP BG, mfss BP BG; BDBCBCBN BCBCBC random edges considered for hillclimbing) dataset rand hlclmb CBBUC6CB CBBUC6CB+MIe CBBUC6CB+MIe+2nd CBBUC6CB+MIe+2nd+hlclmb

in Tractable learning of large bayes net structures from sparse data
by Anna Goldenberg 2004
"... In PAGE 5: ...eported in terms of the average BDeu score, i.e. the fi- nal BDeu score obtained by the network averaged over the number of records in the dataset. The number of edges in the resulting Bayes Nets is reported in Table5 . It is interest- ing to note that the BDeu scores correspondingto the Bayes Nets obtained by running CBBUC6CB as described in Table 1 are very close to the ones obtained by random hillclimbing, but have significantly lower number of edges.... ..."
Cited by 19

Table 1: The estimated and actual times of two programs under differing axis and stride alignments. The times measured were averaged over ten runs. The solution reported for the random edge selection heuristic reflects the best of five trials. The table shows the estimated cost according to the discrete metric and the actual execution time of the program. Example Method Comm. CM5 running

in Algorithms for Automatic Alignment of Arrays
by Siddhartha Chatterjee, John R. Gilbert, Leonid Oliker, Robert Schreiber, Thomas J. Sheffler
"... In PAGE 18: ...3.2 Execution time We show in Table1 the execution time of each program on the CM-5 with each alignment our algorithm generated, and also without axis or stride optimization as a baseline. We draw two conclusions.... ..."

Table 9.1: Experimental results. It can be seen, that if the iteration count is increased, the num- ber of needed octree cubes grows much slower than the number of corresponding voxels. Note also, that the reconstruction of the more complex upper part of the roots (Fig. 9.2) needs more time and more cubes for reconstruction than the simpler lower part (Fig. 9.1 d), when using the same refinement level. Note also the effect of the artefacts inside Fig. 9.1 a. Reconstruction was done at 1.2 GHz CPU speed.

in Three-Dimensional Volume Reconstruction of Rice Roots Filip Sadlo
by Diploma Thesis September, Filip Sadlo 2003

Table 1: Comparisons of adaptive solid texturing and two previous methods. During the rendering state, the 3D-texture mapping and using single cache cube in our approach could achieve a real-time response, but the traditional solid texturing could not. To get a similar quality of the previous solid texturing, we use multiple cache cubes with LOD control- ling. Although the performance is a little slower then us- ing a single cache cube, but the appearance result is much better than using a low resolution one. The result of viewing the inner texture is shown in Figure 2 left. Because our method does not only simulate the surface of the object, but also reserves a space for stor- ing the interior texture data, showing the inner texture of

in Adaptive Solid Texturing for Web3D Applications
by unknown authors

Table II. Cubes Were Larger than the Sphere; Error Tolerance Is the Length of the Cube Minus the Diameter of the Sphere

in PRISM Interaction for Enhancing Control in Immersive Virtual Environments
by Scott Frees, G. Drew Kessler, Edwin Kay

Table 2: RMS Errors of The Cube

in unknown title
by unknown authors
"... In PAGE 6: ...0 pixel widths. Table2 shows the RMS error after plane #0Cttings. We observe that both MFS1 and MFS have better per- formance than KOA, a nd the RMS error of MFS2 re- mains almost constant when the surface slant increases whichwe attribute to the slope correction performed by MFS2, while the error of KOA increases obviously as the slope increases.... ..."

Table 2 The cube and the loop identities of residual systems

in Four Equivalent Equivalences of Reductions
by Vincent Van Oostrom, Roel De Vrijer
"... In PAGE 28: ... A residuation will also be called a residual or projection operation. The flrst identity of Table2 is called the cube identity and the other identities the loop identities (see Figure 16 for a two-dimensional rendering of the cube identity). Notions for abstract rewrite systems extend to residual systems via their underlying abstract rewrite systems.... ..."

Table 6. Abstraction functions

in IOS Press Abstract Interpretation against Races
by Roberto Barbuti, Stefano Cataudella, Luca Tesei
"... In PAGE 8: ... In particular we define an abstract function for each syntactic category, thus we define AB D6 BM D9 AX D9 CL , AB CS BM CS AX CS CL , and so on. The definition of AB functions is given in Table6 where, for the sake of readability, the argument of each function is an element of the concrete syntactic categories rather than a set. The abstraction of a set of concrete elements is defined, as usual, as the least upper bound of the abstractions of the single elements of the set.... ..."

Table 6. Nile performances. Without accessors, Nile is faster than Squeak. But using them makes it slower.

in Redesigning with Traits: the Nile Stream trait-based Library
by Damien Cassou, Stéphane Ducasse, Roel Wuyts
"... In PAGE 19: ... As Squeak does not have a JIT compiler, using accessors instead of direct instance variable access has a cost. Table6 shows that using accessors in the context of stream on strings and arrays is 41% times slower than direct instance variable access. To optimize our library as much as possible we used direct accesses, i.... ..."
Results 11 - 20 of 62,290