Toward an Organic Computing Approach to Automated Design of Processing Pipelines

This paper aims to propose a novel Organic Computing concept to dealing with the overall issue of automated design of processing pipelines. It is outlined how several methods standing under the Artificial Intelligence umbrella are combined to form a technique that can be realized by Organic Computing systems to strengthen their self-configuration property by implementing self-optimization and self-learning techniques. Three envisioned application scenarios are discussed which will serve as first testbeds for the proposed architecture in a future research project: The automated design of 1) a data pre-processing observer component for the refurbishment and the analysis of insufficient quality data to improve the learning ability of employed machine learning algorithms, 2) an image processing pipeline for industrial imaging systems, and, 3) production lines in manufacturing scenarios.