ACTIVE LEARNING AGENT FOR LEARNING AT LEAST ONE STRATEGY

The invention is directed Active learning agent (1) for learning at least one strategy, comprising a. an input interface (10), configured for receiving an input data set; comprising a plurality of labelled sample pairs and a plurality of unlabelled sample pairs; wherein the input data set is associated with a semi-supervised machine learning model; b.a transforming unit, configured for transforming the received input data set into a transformed input data set in form of a transition matrix or knowledge graph; and c.an active learning unit (20), configured for learning at least one strategy based on the transformed input data set using reinforcement learning; wherein the at least one strategy is a strategy for selecting an output subset of the plurality of unlabelled sample pairs of the input data set; and d. an output interface, configured for providing the output subset. Further, the invention relates to a technical system.