### Table 1: Non-optimized network for order 4 Number of generated

"... In PAGE 5: ...umeric (e.g. type of customer = 1,2 or 3) or string (e.g. sex = M or F). To better evaluate the improvements of the optimized algorithm, 3 tests were performed: 1. A complete CNM network is generated ( Table1 ). This network contains all combinations for the specified 32 findings from order 2 to order 4.... In PAGE 7: ... The columns Remaining rewarded combinations and Final number of combinations should have the same values for the three algorithm versions. The first test ( Table1 ) has different values due to remaining combinations among findings of the same fuzzy evidence. Those combinations should be eliminated once two different values of one evidence must never happen.... In PAGE 8: ... The optimized network (Table 3) used only 43.38% of the memory used by the non-optimized ( Table1 ). Because of such memory reductions during the learning, it is possible to generate the neural network up to order 5 using the optimization algorithm.... In PAGE 8: ... For the optimized algorithm (Table 3), it is necessary to add only the Remaining rewarded combinations of previous orders, and the column Number of generated combinations for order for the order in learning process. Considering learning order 4, for the non-optimized network ( Table1 ) the total of combinations is 82832 while for the optimized network (Table 3) is 29111, that means 64.86% of reduction.... ..."

### Table 1: BAT Results with Non-Optimized Ordering

2002

"... In PAGE 9: ... With the manual effort try- ing to interleave the symbolic vectors, symbolic simulation can handle up to 7 bits of symbolic values. Table1 shows the run times for using from 1 to 7 symbolic bits in ea and BEPI.If more than 7 symbolic bits are used, then the run time would take too long.... In PAGE 9: ... In this sequence of experiments, we were able to complete the assertion check without using any constant logic values to simplify the assertion. By comparing the results in Table1 and in Table 2, we observe that variable ordering sig- nificantly impacts the performance of symbolic simulation. For the OBDD sizes, we show two types of data: the total number of OBDD nodes at the end of symbolic simulation (to- tal OBDD nodes), and the maximum number of OBDD nodes during the symbolic simulation (max OBDD nodes).... ..."

Cited by 1

### Table 5: Approximations and non-optimal pricing schemes.

2006

"... In PAGE 24: ... instance, in the second case, (1; 0:90; 0:90) indicates full price in period 1, and 10% discount in the remaining two periods.) The results are summarized in Table5 , where the last column is the percentage o the objective value under optimal pricing. From the above results, we observe that (a) reducing the number of price markdowns from 7 (m = 8) to 2 (m = 3) has a rather minor e ect on the objective values; (b) with more inventory available for sale, price reduction becomes more substantial and starts earlier, as expected; (c) the approximation scheme in (53) performs quite well in all three cases; (d) applying the optimal pricing results in a substantial advantage over other ad-hoc schemes.... ..."

### Table 5: Approximations and non-optimal pricing schemes.

"... In PAGE 23: ... We take the above cases under m = 3, and compute the approximations following (45); we also examine a set of alternatives that ofier difierent levels of discount at difierent periods or no discount at all. The results are summarized in Table5 , where the last column is the percentage ofi the optimal objective values (in Table 3). From the above results, we observe that... ..."

Cited by 1

### Table 3 Clustering measures of 14 data sets using the non-optimal parameter setting; average values range from 0-1 where 1 represents the best results

2001

"... In PAGE 7: ...2. Non-Optimal Parameters Table3 gives the results using the second parameter set. These parameters were obtained in the Genetic Algorithm where the fitness function deliberately restricted the size of the tree for the CENT II model, which means that the trees produced using this parameter set may be non-optimal for data with many clusters.... ..."

Cited by 1

### Table 4: Non-Optimal Portfolio of Demand Manage- ment Contracts with Fixed =0:7 Customer Amount Curtailed Incentive O ered

1999

"... In PAGE 5: ...xed to be 0.7, i.e. the utility decides that the value of power interruptibility for each customer is the same. As shown in Table4 the number of participating customers increases but the amount of available relief decreased and the result was also a smaller increase in the load- ing margin. In the other simulation the utility assumes that the costs of an outage is the same to all customers 4In order to convert from normalized numbers base for the amount of agreed relief of 10 MW and base for the incentive of... ..."

Cited by 1

### Table 12: Non-optimized router interface

1997

Cited by 1

### Table 1. Non-Optimized Optimized

"... In PAGE 9: ... Table1 : Statistics *For the purpose of these statistics, one range check is treated as two range-check compares. This is done because the compiler is sometimes able to eliminate just the low half or the high half of a range check, and this is a way to incorporate the half-checks into the statistics.... ..."