Interactive Machine Learning with Evolving Fuzzy Systems
The central goal of this project is to develop methods enabling advanced interaction (and
communication) between humans and evolving soft computing models (especially Evolving (neuro-)fuzzy systems) in an ongoing on-line, incremental learning context, where knowledge changes, expansion, and contraction in the model(s) are realized by an on-going adaptation with both, data and human input (=> hybrid modeling paradigm).
This should be ideally constituted in a way such that the communication is as economic and as
efficient as possible for human users and operators—ideally only in particular situations, humans should be requested for feedback by the system.
Interpretability of the models and their outputs will be a necessary aspect in order to stimulate any interaction with users at all --- hence, E(N)FS provides an adequate model architecture, but several techniques for improved interpretability of EFS will be needed. A feedback integration component will be essential in order to properly integrate human feedback into the evolving models trained from data.
None, Basic Research Project (FWF)