"... In PAGE 3: ... They are not meant to be optimal. Table1 summarises the results of 25 independent runs of our algorithm at epoch 500 and 1000 for the Aus- tralian credit card assessment problem. Figure 1 shows the average result over the 25 runs.... In PAGE 4: ... It can produce neural network ensembles with good generalisation abil- ity without using techniques such as cross-validation. The results given in Table1 is comparable to those obtained using evolutionary approaches [8, 9], but bet- ter than others [6]. Table 3 compares the average testing error rates among di erent algorithms.... ..."

"... In PAGE 6: ...1 Credit Screening Data In the credit screening, each of five datasets contains 523 samples in the training set and 130 samples in the testing set. Table5 shows the classification results on the training and testing sets. The average correct classification rates are 88.... ..."

"... In PAGE 8: ...8 Anecdotally, middle market lenders in all three countries have identified this apos;gaming apos; of the system as potentially relevant, and no more or less so in any of them. Table3 illustrates two indicators of bank credit quality that help guide our top-down opinion of aggregate default probabilities within these countries. First, in general the banking sectors in the various countries have roughly similar ratios of nonperforming loans as a percent of total assets (column 3), which implies that asset quality is approximately equal across the countries.... In PAGE 8: ... The different proportions of con- sumer and small business lending in these figures clearly make aggregate comparisons difficult, nonethe- less they are still informative. Another top-down note within Table3 is the median P/E of banks, especial- ly the P/E ratio of banks relative to other corporates in each country. The ratio is highest in Australia, then Canada, and lastly the US.... ..."

"... In PAGE 26: ... We chose 55 in order to obtain clean coefficients on the maturity variable, as explained in Appendix C. 24 The rescheduling variable appears in Table4 for low-rated issuers but not in Table 5 for high- rated issuers because there are no highly-rated countries with recent histories of rescheduling in the sample, not surprisingly given the way the rating agencies operate. 25 Thus, while the results in the final columns of Tables 2-5 include data through 1999-IV (given the difficulty of estimating the full model on the relatively small number of observations for the sub-period following the Russian crisis), the out-of-sample forecast errors in the final column of Tables 6-8 are based on a model estimated through 1998-IV.... ..."

"... In PAGE 26: ... We chose 55 in order to obtain clean coefficients on the maturity variable, as explained in Appendix C. 24 The rescheduling variable appears in Table 4 for low-rated issuers but not in Table5 for high- rated issuers because there are no highly-rated countries with recent histories of rescheduling in the sample, not surprisingly given the way the rating agencies operate. 25 Thus, while the results in the final columns of Tables 2-5 include data through 1999-IV (given the difficulty of estimating the full model on the relatively small number of observations for the sub-period following the Russian crisis), the out-of-sample forecast errors in the final column of Tables 6-8 are based on a model estimated through 1998-IV.... ..."

"... In PAGE 31: ... It is hard to imagine an explanation for these patterns grounded purely in the technical characteristics of the relevant markets.47 Finally, Table5 reports the results for spreads, separately for countries in different Institutional Investor rating categories. Many of the coefficients are less precisely estimated than on the full set of bonds (as shown in Eichengreen and Mody 2000b), reflecting the now smaller sample size.... In PAGE 34: ...54 Among the objections to this idea, as noted above, is that guarantees may do more to signal problems than to reassure investors. Table5 shows that guarantees reduce spreads on issues from low-rated countries (0-30 on the Institutional Investor scale), as if the bad news about these countries is already known and the main effect of guarantees is to reduce risk. This is the same group of countries, it will be recalled, for which IMF programs have no discernible impact on spreads.... ..."

"... In PAGE 6: ...1 -4.3 Notes: a) minus sign denotes a rise and vice versa b) 1989 to 1996 Source: IMF These capital inflows were correspondingly associated with a sharp rise in broad money (M2) in the SEA economies ( Table3 )1. The incentive for the foreign lending boom seems clear from the significant and sustained interest rate premia offered by the crisis-hit SEA economies over the LIBOR rate, on the one hand, and the stable exchange rate regimes they maintained, on the other (Table 4).... ..."

"... In PAGE 5: ...6 List of Tables and Figures Table 1: Growth rate of GDP components in different periods Table 2: Growth rate of GDP components in different periods Table 3: APP irrigation targets and progress during the ninth plan Table 4: Import of fertilisers by sector and type of fertilisers Table 5: APP agricultural road target (physical and financial) Table 6: Physical and financial targets and progress Table7 : APP estimates of credit requirement Table 8: Lending activities of ADBN on APP specified purposes, 2002/03 and 2003/04 Table 9: Awareness and influence of APP in programmes of GOs in the study districts Figure 1: Devolution structure envisaged by LSGA 1999 ... In PAGE 21: ... However, with the change experienced in the use of STWs, the role of electricity in the context of APP may have also changed. Rural credit supply (main focus on plains - limited attention to mountain/hills) Rural credit supply was important under the APP to enable farmers to access improved agricultural technology ( Table7 ). The Agricultural Development Bank of Nepal (ADB/N) was a major player and has provided almost 50 percent of the annual rural credit needs (Table 8).... ..."