Learning qualitative models from numerical data

Qualitative models are often a useful abstraction of the physical world. Learning qualitative models from numerical data sible way to obtain such an abstraction. We present a new approach to induction of qualitative models from numerical data which is based on discrete Morse theory (DMT). Our algorithm QING (Qualitative INduction Generalized) has a firm theoretical background in computational topology. This makes it possible to extend the capabilities of state-of-the-art algorithms for qualitative modelling substantially. The output of QING is a labeled graph, which enables a visualisation of the qualitative model. Induced qualitative models can also be used for numerical regression by applying the Q2 method. To illustrate the power of QING we present its application on an artificial function, add noise, and finally show how it performs on a dynamic domain such as inverted pendulum.

• Type: Journal
• Year: 2011
• Where: AI
• Authors: Jure Zabkar, Martin Mozina, Ivan Bratko, Janez Demsar
• Keywords: AI 2011, Artificial Intelligence, Numerical Data, QING, Qualitative Models