Figure 2
From: Logic Learning Machine creates explicit and stable rules stratifying neuroblastoma patients
Rules stabilization flowchart. The initial dataset is analyzed by LLM to generate the initial classifier and the stability of the rules is assessed in cross validation as described in the Methods section. The stability of each rule is associated to the classifier which is accepted if the stability of every rule is equal to one. If this condition does not apply, Core rules will be derived from the 50 classifiers originated by the cross validation procedure. These rules allow to identify the instances that generate instability which are deleted from the dataset. The purged dataset serves as input of a new cycle of analysis which is repeated till each rule has stability equal to one.