Abstract: In this ,paper, it is assumed that a point at which a unimodal func-
tion possesses a maximum value satisfies a given probability distribution,
For a giveen policy, we have defined a sequential decision _procss, Given
numbers s and d, we have discussed the optimality of the policy ,when
mathematical expectation of the number of tests according to the policy
s-been taken as an optimum criterion, and obtained a necessary and su-
fficient condition that an optimal policy must satisfy, Meanwhile, we
have discussed the optimization about nn-step policy when mathematical
expectation of final 8-accuracy has been taken as an optimum criterion,
and finally obtained similar results,

Key words: optimum seeking mothod, probability, optimal policy, conditions