An evaluation ofalgol 68 for interrogation process algorithms

Abstract

Given a random processT taking a finite number of statesT _j on a population Ω, we consider a family,D, of random processes defined on Ω. Each elementq ofD has a finite number of possible statesq _i, and we suppose that the conditional probabilitiesp(q _i|T _j) are known. We consider arborescent decision processes using the elements ofD as questions and aiming at determining the state of the systemT during a given experiment ω of ω. The questionnaires of Picard treat the cases where all the conditional probabilities are equal to 0 or 1; we introduce the pseudoquestionnaires to handle the general case. Certain recognition problems of structure can also lead to the construction of a pseudoquestionnaire. We will show here in an example that algol 68 appears to be well suited to the writing of such algorithms.