"... In PAGE 6: ....1. Experimental data The proposed method has been tested and compared with the conventional ones. This was done by performing experi- ments on a number of data sets as summarized in Table1 . The sample vectors of each data set are divided into two subsets of equal size, and used for training and validation, alternately.... ..."

"... In PAGE 16: ...Table 2: Real-life Data Sets We experimented with two real-life data sets whose characteristics are illustrated in Table2 . These data sets were obtained from the UCI Machine Learning Repository5.... In PAGE 16: ... For these data sets, we randomly choose 2/3 of the data and used it as the training data set. The last column in Table2 contains the number of nodes in the nal tree constructed by PUBLIC without constraint pruning. 5.... ..."

"... In PAGE 8: ... Thus if d gt; c log2 n, we have p lt; 1 !clog2 n = n?c log2(1= ) = n?0:693c: To give a numeric example, in Section 4 below one of the Hu man trees corresponds to the di erent words in English, with n = 289; 101 leaves. The probability for a tree of this size to have a leaf at level 3 log2 n is less than 4:4 10?12, meaning that such a word occurs only once every 4400 billion words; the existence of such a rare word then puts a lower limit on the size of the text, which in our case must be large enough to ll about 35,000 CD-Roms! For all the distributions given in Table2 in the experiments below, the ratio of the depth of the Hu man tree to log2 n is between 1.... In PAGE 10: ...Table 2: Time and Space requirements for real-life distributions Table2 displays the results. The rst three columns give some statistics about the various distributions: the depth k of the Hu man tree, the size n of the encoded alphabet, and the weighted average length of a codeword, measured in bits, which equals the average number of comparisons if the standard Hu man tree is used.... In PAGE 15: ...But with standard binary search, the search in a code with maximal codeword length k takes exactly dlog2 ke comparisons, which would be 4 or 5 for our example distributions. Note that while the average number of comparisons with sk1-trees is above that threshold for all the examples ( Table2 ), all the corresponding values for the sk2-trees are below it (Table 3). It does therefore not seem necessarily worthwhile to pass to skd-trees, for d gt; 2.... ..."

"... In PAGE 7: ... Function Gen- erator describes a ramp-signal generator [6]. Table1 presents the main results of the synthesis experi- ments. All examples were synthesized so that the global area was minimized.... ..."

"... In PAGE 5: ... For the experimental study we used nine real life and three synthetic datasets. Their characteristics are summarized in Table1 . All datasets except 3DSin have been used before extensively in experimental studies.... ..."

"... In PAGE 6: ... The databases and their descriptions are available on the UCI Machine Learning Repository [13]. The number of rows, columns, and minimal dependencies found (N) in each database are shown in Table1 . The datasets labeled Wisconsin breast cancer n are concatenations of n copies of the Wisconsin breast cancer data.... In PAGE 6: ... To avoid duplicate rows, all values in each copy were appended with a unique string specific to that copy. The top three rows of Table1 show the performance of the algorithms on three small databases. Our algorithms perform competitivelyin all cases.... In PAGE 7: ... This is a good demon- stration of how different approaches to pruning the search space have different effects. The bottom part of Table1 reports the performance of TANE on five large databases. For TANE/MEM and FDEP, some experiments are marked with (*) as infeasible; for TANE/MEM because of the lack of main memory, and for FDEP if it did not finish within 5 hours.... ..."

"... In PAGE 14: ... Table1 : SASO-ST Average of Subject Test Ratings As mentioned earlier there are various trade offs and issues with the technology that need to be addressed. These virtual human systems can be built and fielded today, however it requires several experts to make them useful.... ..."