Use of reliability engineering concepts in machine learning for classification

In reliability engineering, the reliability of a system is estimated by considering the dependencies between the system's components. The probability of a system failure is then expressed in terms of the states of its components. Meanwhile, in some machine learning approaches, the probability of class membership is expressed in terms of the values (which can be seen as `states') of various features (which can be seen as `components'). In this paper, we explore this analogy further to develop a classification algorithm in which the decision for class membership is based on specific combinations of feature states, where those combinations are inspired by reliability engineering. In essence, our classification model considers the features to be components arranged in a parallel in a system, while different classes represent different system configurations. We describe the approach, present initial promising results, and speculate on further development of the approach.