Learning from errors: A bio-inspired approach for hypothesis-based machine learning

This contribution present an approach extending existing learning strategies based on situation-operator-modeling (SOM), which can be used to model interactions with the environment and to represent the knowledge of cognitive systems. The approach proposes a planning process using hypotheses to bridge the gap of knowledge, which is refined by a following check of the applied hypothesis. The hypotheses are inspired by human errors according to Dornerpsilas classification, which is related to the interaction within complex dynamic systems. The programmed implementation of the approach is based on an experimental environment using a software tool for high-level Petri nets.