Incorporando Aprendizado Incremental com MaxSAT em um modelo de Aprendizado de Regras baseado em SAT

  • Antônio Carlos Souza F. Júnior Instituto Federal de Educação, Ciência e Tecnologia do Ceará
  • Thiago Alves Rocha Instituto Federal de Educação, Ciência e Tecnologia do Ceará
Keywords: Interpretable Artificial Intelligence, Explainable Artificial Intelligence, Machine Learning, Computational Logic, Satisfiability


This article aims to describe a new incremental model for learning interpretable rules based on MaxSAT, called IKKRR. This new model was based on two other approaches, one based on SAT and the other on MaxSAT. The one based on MaxSAT, called IMLI, presents a technique to increase performance that consists in learning a set of rules by applying the model in a dataset incrementally. This work shows which adaptations were necessary for a model based on SAT to be transformed into one based on MaxSAT. It also shows what was done to make it incremental. Finally, IKKRR and IMLI are compared using diverse datasets. Despite having a smaller number of variables, the IKKRR obtained results comparable to the IMLI.