On Active Learning Strategies for Sequential Diagnosis

When diagnosing a faulty system one is often confronted with a large number of possible fault hypotheses. Sequential Diagnosis (SD) techniques aim at the localization or identification of the ac- tual fault with minimal cost or effort. SD can be viewed as an Active Learning (AL) task where the learner, trying to find some target hypothesis, formulates sequential queries to some oracle, thereby e.g. requesting additional system measurements. Several query selection measures (QSMs) for de- termining the best query to ask next have been proposed for AL. To date, few of them have been translated to and employed in SD. In this work, we account for this and analyze various QSMs wrt. to the discrimination power of their selected queries within the diagnostic hypotheses space. As a result, we derive superiority and equivalence relations between these QSMs and introduce improved versions of existing QSMs to overcome identified issues. The obtained picture gives a hint about which QSMs should preferably be used in SD to choose a query from a pool of candidates. Moreover, we deduce properties optimal queries wrt. QSMs must satisfy. Based on these, we devise an efficient heuristic search for optimal queries. As (preliminary) evaluation results indicate, the latter is especially beneficial in applications where query generation is costly, e.g. involving logical reasoning, and hence a pool of query candidates is not (cheaply) available.