There are many real scenarios in which some correlated complex problems have to be addressed by different autonomous learners working in parallel. In such a scenario, the collaboration among the learners can be extremely useful since they can share acquired knowledge so as to reach a reduction in the learning time, an increase in the learning quality, or both of them. Anyway, in some cases, it is not always feasible to collaborate with other learners. This is because the problems to solve are not compatible or they can have dissimilar boundary conditions leading to very different problem solutions. In this paper, we propose an approach to collaborative learning which leverages cellular automata for efficiently solving a set of compatible and sufficiently similar problems. In this direction, the notion of compatibility and similarity between problems is also given and discussed. A case study based on the maze problem will show the effectiveness of the proposed approach.KeywordsCollaborative LearningCellular AutomataArtificial IntelligenceReinforcement Learning
Collaborative Learning over Cellular Automata
Emilio Greco;Antonio Guerrieri;Giandomenico Spezzano;Andrea Vinci
2023
Abstract
There are many real scenarios in which some correlated complex problems have to be addressed by different autonomous learners working in parallel. In such a scenario, the collaboration among the learners can be extremely useful since they can share acquired knowledge so as to reach a reduction in the learning time, an increase in the learning quality, or both of them. Anyway, in some cases, it is not always feasible to collaborate with other learners. This is because the problems to solve are not compatible or they can have dissimilar boundary conditions leading to very different problem solutions. In this paper, we propose an approach to collaborative learning which leverages cellular automata for efficiently solving a set of compatible and sufficiently similar problems. In this direction, the notion of compatibility and similarity between problems is also given and discussed. A case study based on the maze problem will show the effectiveness of the proposed approach.KeywordsCollaborative LearningCellular AutomataArtificial IntelligenceReinforcement LearningI documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.