The minimax approach to the construction of a model class software reliability based on a combination of generalized bayesian inference, and elements of learning theory

The paper proposes a new approach to constructing models of software reliability, using a combination of generalized Bayesian inference, and elements of learning theory. To construct the models we consider a set of probability distributions, reflecting the lack of available statistical data, and on this set, search for the maximum of functional risk is done. The model parameters are computed by minimizing the obtained functional. This approach is quite general and allows to develop various models of reliability.