### Table 1: Performance for prediction 6 steps ahead (exp A.) and 84 steps ahead (exp B.) The results from previous works were taken from [4]. Method Training NDEI NDEI

### Table 1: Mean estimated return and standard deviation for merging and jumping and comparable work in the literature. VAPS is detailed in [11], Exp-GPOMDP, IState- GPOMDP, and GAMP in [1]. Note that results for VAPS are estimated from graphs.

### TABLE II. Physical properties of diphenyl carbonate. Units are: density, , kg m?3; heat of vaporization, Hvap, kJ mol?1; di usion coe cient, D, 10?5cm2s?1. `time apos;=length of trajectory used for averaging. Experimental sources: density, heat of vaporization from Ref. [19]. Subscripts: FF(1), FF(2), simulation with force elds of this work; exp, experiment. T /K 350 393.15 400 450 500 573.15 600

in Structure and Dynamics of Liquid Diphenyl Carbonate Investigated By Molecular Dynamics Simulations

### Table 2: Results of the experiment no. 1, 2 and 3. exp. no.1 exp. no.2 exp. no.3

"... In PAGE 4: ... In the experiment no.1, we gathered images from the Web for 20 kinds of class keywords shown in Table2 . By the image-gathering module about ten thousands URLs were fetched from three commercial text search engines, Google, InfoSeek, Goo Japan.... In PAGE 4: ...8%, which is defined to be NOK/(NOK + NNG), where NOK, NNG are the number of relevant images and the num- ber of irrelevant images to their keywords. In the second and third columns of Table2 , we show the number of URLs of images gathered from the Web and their precision. In the columns after the fourth of Table 2, we show the classification result using the gathered images from the Web as training images.... In PAGE 4: ... In the second and third columns of Table 2, we show the number of URLs of images gathered from the Web and their precision. In the columns after the fourth of Table2 , we show the classification result using the gathered images from the Web as training images. Note that the precision of train- ing images is not 100% unlike the conventional works on image classification.... In PAGE 4: ...3 as the F-measure value by color signatures. In Table2 , we also show the results of the experiment no.... ..."

### Table 5 shows number of statements in standard and reduced ontologies for random ontologies generated for different values of conceptCount, expCount, expMaxSize. As can be seen in this table, the number of statements is re- duced after applying the reduction algorithm on these ontologies. This shows that, SBAC needs to work with smaller ontologies and therefore it requires a lower space capacity.

"... In PAGE 14: ... Table5 . Number of statements in standard and reduced ontologies conceptCount expCount expMaxSize Statements of Statements of Standard Ontology Reduced Ontology 100 20 10 390 239 1000 20 10 1262 1130 1000 100 10 2382 1688 500 200 10 3158 1825 1000 200 10 3600 2300 5000 500 20 16194 10593 Table 6.... ..."

### Table 5 shows number of statements in standard and reduced ontologies for random ontologies generated for different values of conceptCount, expCount, expMaxSize. As can be seen in this table, the number of statements is re- duced after applying the reduction algorithm on these ontologies. This shows that, SBAC needs to work with smaller ontologies and therefore it requires a lower space capacity.

in Preface

"... In PAGE 51: ... Table5 . Number of statements in standard and reduced ontologies conceptCount expCount expMaxSize Statements of Statements of Standard Ontology Reduced Ontology 100 20 10 390 239 1000 20 10 1262 1130 1000 100 10 2382 1688 500 200 10 3158 1825 1000 200 10 3600 2300 5000 500 20 16194 10593 Table 6.... ..."

### Table 3: Precision at ranks on the working set.

2005

"... In PAGE 5: ...Table 3: Precision at ranks on the working set. In Figure 3 and Table3 , we show the precisions at di erent ranks on the working set for the four methods. We see that the baseline method ExpCorr performs the worst, while BM25Corr performs the best.... In PAGE 5: ... This means that the performance we see actually represents a lower bound; the real performance can only be better. Comparing Table3 and Table 4 indicates that the baseline ExpCorr performs similarly, indicating that the additional 30 Chinese documents retrieved mostly have not made to the top pairs. The slight decrease in the precision at rank 50 and rank 100 suggests that there may be a couple un- judged Chinese documents showing up in the top 100 list.... ..."

Cited by 1

### Table 1: Best results obtained with training data CLEF 2006 language sp wm C avgld exp np nd avgP

in General

"... In PAGE 3: ... This value was obtained with the 2006 collections. Table1 shows the best configuration for each language: Table 1: Best results obtained with training data CLEF 2006 language sp wm C avgld exp np nd avgP... In PAGE 3: ...09% for Hungarian. 5 Conclusions and Future Work In this eighth CLEF evaluation campaign, we compared different query expansion techniques in our system for Hungarian, Bulgarian and Czech (see Table1... ..."

### Table 2. Expected work for ECM

1996

"... In PAGE 14: ...n two variables. Suppose that the minimum is Wopt. Tables of optimal parameters are given in [3, 7, 54, 59, 79], with each paper making slightly di erent assumptions. In Table2 we give a small table of log10 Wopt for factors of D decimal digits. We assume that K1 = 11= log 2, K2 = 1, and log10 p apos; D ? 0:5.... In PAGE 14: ... We assume that K1 = 11= log 2, K2 = 1, and log10 p apos; D ? 0:5. Some computed values of (p) are also shown in Table2 , where (p) = (log Wopt)2 log p log log p ; so Wopt = exp q (p) log p log log p : Since the expected run time is insensitive to changes in B1 and B2 near the optimal values, it is not important to choose them accurately. In practice, the signi cant point is that we do not know p in advance.... In PAGE 15: ... Thus, we expect (p) ! 2 as p ! 1, independent of whether phase 2 is used. Table2 shows that the convergence is very slow and that (p) apos; 1:7 for p in the 25 to 45 digit range. We note a lack of symmetry in (29) which may be of interest to cryptographers [70].... In PAGE 19: ...actor N. Each curve takes about 22:9 B1 such multiplications. Overall, our factorization of F10 took 1:4 1011 multiplications (mod N), where N = c291. ( Table2 predicts 3:3 1011 with the optimal choice of parameters.) Numbers mod c291 were represented with 38 digits and base 226 (on the VP100/VP2200) or with 41 digits and base 224 (on the Sparc), so each multiplication (mod N) took more than 104 oating-point operations.... In PAGE 24: ...actor can be obtained from Vershik apos;s result [81, Thm. 3], which is paraphrased above. On the other hand, Harvey Dubner and the author have tried more than 500 curves with B1 = 106, in an attempt to factor F12, without nding more than the ve known prime factors. Thus, from Table2 and (29), we can be reasonably con dent that the sixth-smallest prime factor of F12 is at least 1030; a smaller factor would have been found with probability greater than 0:9. The complete factorization of F12 may have to wait for the physical construction of a quan- tum computer capable of running Shor apos;s algorithm [76], or a surprising new development such as a classical (deterministic or random) polynomial-time integer factoring algorithm.... ..."

Cited by 16

### Table 2. Expected work for ECM

1996

"... In PAGE 14: ...n two variables. Suppose that the minimum is Wopt. Tables of optimal parameters are given in [3, 7, 54, 59, 79], with each paper making slightly di erent assumptions. In Table2 we give a small table of log10 Wopt for factors of D decimal digits. We assume that K1 = 11= log 2, K2 = 1, and log10 p apos; D ? 0:5.... In PAGE 14: ... We assume that K1 = 11= log 2, K2 = 1, and log10 p apos; D ? 0:5. Some computed values of (p) are also shown in Table2 , where (p) = (log Wopt)2 log p log log p ; so Wopt = exp q (p) log p log log p : Since the expected run time is insensitive to changes in B1 and B2 near the optimal values, it is not important to choose them accurately. In practice, the signi cant point is that we do not know p in advance.... In PAGE 15: ... Thus, we expect (p) ! 2 as p ! 1, independent of whether phase 2 is used. Table2 shows that the convergence is very slow and that (p) apos; 1:7 for p in the 25 to 45 digit range. We note a lack of symmetry in (29) which may be of interest to cryptographers [70].... In PAGE 19: ...actor N. Each curve takes about 22:9 B1 such multiplications. Overall, our factorization of F10 took 1:4 1011 multiplications (mod N), where N = c291. ( Table2 predicts 3:3 1011 with the optimal choice of parameters.) Numbers mod c291 were represented with 38 digits and base 226 (on the VP100/VP2200) or with 41 digits and base 224 (on the Sparc), so each multiplication (mod N) took more than 104 oating-point operations.... In PAGE 24: ...actor can be obtained from Vershik apos;s result [81, Thm. 3], which is paraphrased above. On the other hand, Harvey Dubner and the author have tried more than 500 curves with B1 = 106, in an attempt to factor F12, without nding more than the ve known prime factors. Thus, from Table2 and (29), we can be reasonably con dent that the sixth-smallest prime factor of F12 is at least 1030; a smaller factor would have been found with probability greater than 0:9. The complete factorization of F12 may have to wait for the physical construction of a quan- tum computer capable of running Shor apos;s algorithm [76], or a surprising new development such as a classical (deterministic or random) polynomial-time integer factoring algorithm.... ..."

Cited by 16