CiteSeerX — Table search results

Results 1 - 10 of 4,081

Table 1: Domains for which the greedy on-line meta-deliberation procedure is optimal. Key: a107

in Research Note On the Optimality of Greedy Meta-Deliberation

by David C. Parkes, Lloyd G. Greenwald

"... In PAGE 5: ... What is required is information about the distribution over challenge types, given that some challenge has occurred. Table1 characterizes the domains for which the greedy procedure is optimal. It is optimal for all challenges and environments with linear-increasing (weakly-concave), or concave-increasing value-of-precomputation functions.... ..."

Table 3: On-line performance of GREEDY and GREEDYT. For each, Time is the time in seconds to compute the partition and compress the file; /Gzip is the time relative to gzip. Gzip GREEDY GREEDYT

in Improving Table Compression with Combinatorial Optimization

by Adam L. Buchsbaum, et al. 2002

"... In PAGE 9: ... Table 2 shows the resulting compressed sizes using partitions computed with GREEDY and GREEDYT. Table3 gives the time results. GREEDY compresses to within 2% of DP on seven of the files, including four of the genetic files.... ..."

Table 3: On-line performance of GREEDY and GREEDYT. For each, Time is the time in seconds to compute the partition and compress the le; /Gzip is the time relative to gzip. Gzip GREEDY GREEDYT

in Improving Table Compression with Combinatorial Optimization

by Adam L. Buchsbaum, Glenn S. Fowler, Raffaele Giancarlo

Table 2: Comparing the average results for the greedy online algorithm from 1000 trials per experi- ment with a numerical evaluation from the differential equations (to four decimal places).

in Using the Power of Two Choices to Improve Bloom Filters

by Steve Lumetta, Michael Mitzenmacher

Cited by 1

Table 1: Average results for the greedy online algorithm from 1000 trials per experiment. Recall that n is the number of elements, m is the number of bits, c is the number of choices, and k is the number of hash functions. The value of k with the best observed false positive probability is used for each setting of n, m, and c. Note the case c BP 1 corresponds to the standard Bloom filter.

in Using the Power of Two Choices to Improve Bloom Filters

by Steve Lumetta, Michael Mitzenmacher

"... In PAGE 5: ...f choices is greater than two. We compare the results from this equation to our simulations below. Also, note that the form of equation (3) for one group of hash functions, namely dz dt BP ECJB4B1B4kBN1 A0 zB5CL mBPn BP kB41 A0 zB5 mBPn BN can be solved exactly to yield equation (1). Our experimental results for the online setting are given in Table1 . We have considered cases with 8, 16, and 32 bits per set element, which are standard configurations.... In PAGE 6: ...verage of 3.954e-04, and using 14 gave an average of 4.035e-04. It is also worth noting that the differential equation (3) is very accurate, as we show in Table 2. For the optimal configurations presented in Table1 , the predicted fraction of ones in the Bloom filter from equation (3) matches the average from our experiments to essentially four decimal places (with one case off in the last digit due to rounding). We conclude that equation (3) can be used to accurately predict performance.... ..."

Cited by 1