|
|
|
|
|
function is fed to the scientist in canonical order. Also observe that the auc measure is defined for a scientist on a function if the scientist converges on the function; it does not matter to what the scientist converges. |
|
|
|
|
|
|
|
|
Show that auc satisfies the axioms given in in Definition 12.10. |
|
|
|
|
|
|
|
|
This measure is developed in greater detail in Daley and Smith [54]. |
|
|
|
|
|
|
|
|
12-7 (Freivolds [61], Kinber [106]) The following provides the function analog of the TxtMin paradigm. |
|
|
|
|
|
|
|
|
12.27 Definition |
|
|
|
|
|
|
|
|
(a) We say that M Min y -identifies f (written: ) just in case . |
|
|
|
|
|
|
|
|
(b) . |
|
|
|
|
|
|
|
|
Show that there is an acceptable programming system y for which Min y has an infinite r.e. class of functions as an element. |
|
|
|
|
|
|
|
|
12-8 (Freivolds [61]) Freivolds introduced the notion of limit standardizability with a recursive estimate (abbreviated: Lsr) defined below: |
|
|
|
|
|
|
|
|
12.28 Definition We say that S is limiting standardizable via g with recursive estimate v (written: ) if and only if, for each , there is an if with j if = f such that, for every i with j i = f we have: |
|
|
|
|
|
|
|
|
(a) , and |
|
|
|
|
|
|
|
|
(b) . |
|
|
|
|
|
|
|
|
We say that S is limiting standardizable with a recursive estimate (written: ) if and only if, for some recursive g and v, . |
|
|
|
|
|
|
|
|
(a) Prove that the notion of Lsr is independent of the choice of an acceptable programming system. |
|
|
|
|
|
|
|
|
(b) Establish the following characterization of Min y -identification. The following are equivalent for allS. |
|
|
|
|
|
|
|
|
(i) For some acceptable y , . |
|
|
|
|
|
|
|
|
(ii) . |
|
|
|
|