Extended kernel subset analysis for qualitative model learning

In this paper we continue our previous research on kernel subset analysis for Qualitative Model Learning (QML).We focus on investigating the kernel subsets and learning precision of QML when the number of the training data is relatively large, which makes the corresponding kernel subset experiments very computationally expensive to perform. We use a two-compartment model with two qualitatively different inputs as our testbed to exhaustively perform the kernel subset experiments by the GENMODEL algorithm. An analysis on the obtained experimental results indicates that there exist patterns in the formation of kernel subsets, and the solution space analysis further reveals the distribution of kernel subsets in the solution space.