|
|
|
|
|
Replacing d-conformity by uniform d-conformity in Definition 10.30 produces the following paradigm. |
|
|
|
|
|
|
|
|
10.31 Definition Let and . |
|
|
|
|
|
|
|
|
(a) A scientist M UapdExa-identifies f (written: ) just in case, for all p such that j p is uniformly d-conforming with f, and j M(p,f) =a f. |
|
|
|
|
|
|
|
|
(b) . |
|
|
|
|
|
|
|
|
Ap and Uap are abbreviations for 'Approximate partial additional information' and 'Uniform approximate partial additional information,' respectively. The Uap-type additional information is intuitively better than the Ap type, and indeed the results to follow will confirm this. Since any Uapd-type additional information is also an Apd-type additional information, we have the following. |
|
|
|
|
|
|
|
|
10.32 Proposition For all and all , . |
|
|
|
|
|
|
|
|
We next establish that there are collections of functions that can be identified with some Uap-type additional information of nonzero density, but cannot be identified in the presence of the best possible Ap-type additional information even if the the scientist is allowed to converge to a finite variant program. |
|
|
|
|
|
|
|
|
10.33 Proposition Suppose . Then . |
|
|
|
|
|
|
|
|
Proof: Let n0 = 0 and, for each , let n2i + i + 1 and n2i+2 = n2i+1 · 2i. |
|
|
|
|
|
|
|
|
We show that in the next two claims and thus establish the proposition. |
|
|
|
|