An Incremental Learning Algorithm For Non-deterministic Interpretation Transitions

The modeling of gene regulatory networks (GRNs) has become one of the significant topics in the bioinformatics. Nowadays Boolean network (BN) is one of many available models for network inference, it is well established and remains a topic of considerable interest in the field of genetic network inference. There are many works on inference synchronous boolean network but little about asynchronous. Under asynchronous update mode, the state transition is no longer unique. An algorithm was proposed to learn a disjunctive logic program from interpretation transitions(LFDT). It can be used to learn the dynamic of asynchronous BNs whose state transitions are restricted. But LFDT needs all possible successor states of each state. Usually, this nondeterministic state transitions are incomplete over time. Therefore, LFDT needs to be modified in order to incrementally learn from incomplete observed transitions. For this purpose, a method was proposed to revise the learned hypothesis. The method preserver the equivalence under T
^d
p semantic for existing transitions. The algorithm was altered according to the method. The modified algorithm was proofed to be sound and complete.