Learning qualitative metabolic models
George M. Coghill, Simon M. Garrett, Ross D. King · European Conference on Artificial Intelligence · 2004
The ability to learn a model of a system from observations of the system and background knowledge is central to intelligence, and the automation of the process is a key research goal of Artificial Intelligence. We present a model-learning system, developed for application to scientific discovery problems, where the models are scientific hypotheses and the observations are experiments. The learning system, QoPH learns the structural relationships between the observed variables, known to be a hard problem. QoPH has been shown capable of learning models with hidden (unmeasured) variables, under different levels of noise, and from qualitative or quantitative input data.