Results 1 -
8 of
8
Table 3: Results over 10 runs for the Heaven/Hell scenario. values are multiplied by 102.
2002
"... In PAGE 33: ... IState-GPOMDP contains the additional randomness of di erent trajectories taken through the global state space. Table3 shows the results of these experiments. Both algorithms nd the optimum policy, but only GAMP nds it consistently.... ..."
Cited by 4
Table 1. Two opposite (political) visions on NBIC convergence (based on Garreau 2004, Hughes 2004). Scenario Heaven Hell
2006
Table 12. Hetu Dragon Chart converted to Fu Xi Early Heaven Trigram Numbers.
Table 2. Results for the Heaven/Hell scenario over 10 runs. values are multiplied by 102. Bracketed numbers indicate how many runs converged to = 5:0. Alg ave max var secs to = 5
2002
Cited by 8
Table 3. Luoshu Turtle or Magic Square.
"... In PAGE 7: ...Table 2a: Hetu Dragon Writing Round heaven shape 5, 10 in center = Earth Table3 a. Luoshu Turtle Writing (Lo River Chart) Square earth shape Fire 7 2 Wood 8 3 5,10 4 9 Gold 1 6 Water South 4 9 2 East 3 5 7 West 8 1 6 North (Dots replaced with Arabic numerals in both diagrams.... ..."
Table 1: Racial Composition and Mixture and Housing Stock
2001
"... In PAGE 12: ... Almost as a sign from the heavens, the first child born in Columbia was parented by an inter-racial couple. Examining Census Data (shown in Table1 ), the average value of houses owned by African-American and white households differed by less than $4000. However white families paid significantly higher rents than African-American families (almost $100/month), which reflects the large number of subsidized housing units in Columbia.... ..."
Table 4: Precision of attributes extracted with normalized ranking from Web documents (D) versus query logs (Q), expressed as Absolute scores, Relative boosts (Q over D), and Error reduction rates (Q over D)
"... In PAGE 5: ... The differences in quality are particularly high for classes like HeavenlyBody, CarModel, BasicFood, Flower and Mountain. To better quantify the quality gap, the last rows of Table4 show the precision computed as an aver- age over all classes, rather than for each class individually. Consistently over all computed ranks, the precision is about 45% better on average when using queries rather than doc-... In PAGE 6: ... It shows that query logs represent a competitive resource against document collections in class attribute extraction. As an alternative to Table4 , Table 5 illustrates the top attributes extracted from text for a few of the target classes. Documents produce more spurious items, as indicated by a more frequent presence of attributes that are deemed wrong, such as bowl for BasicFood, mg for Drug, or temple for Heav- enlyBody.... ..."
Table 1. Taxonomy of di cult POMDPs Avoiding getting lost Acting when lost
2000
"... In PAGE 3: ...This paper presents a new approximate solution of POMDPs that works well when there is a delayed need to observe (lower left box, Table1 ). The core idea is to de ne a new POMDP, the even-odd POMDP, in which the full state of the environment is observable at all times t where t is even.... In PAGE 11: ... In addition, the value function V 2MDP is an upper bound on the value of the optimal POMDP policy, which is useful for evaluating hand-derived policies. Because the 2MDP method incorporates a 2-step value-of-information com- putation, it is also suitable for solving problems, such as medical diagnosis, where there is an immediate opportunity to observe, and the two hallways problem, where there is an immediate need to observe (see Table1 ). However, the 2MDP method does not provide any solution to problems where there is a delayed opportunity to observe, such as the \heaven and hell quot; problem.... ..."
Cited by 5