The aim of this study is to introduce a fuzzy model to process structured data. A structured organization of information is typically required by symbolic processing. Most connectionist models assume that data are organized in a form of relatively simple structures such as vectors or sequences. In this work, we propose a connectionist model that can directly process labeled trees. The model is based on a new category of logic connectives and logic neurons that use the concept of uninorms. Uninorms are a generalization of t-norms and t-conorms used for aggregating fuzzy sets. Using a back-propagation algorithm we optimize the parameters of the model (relations and membership functions). The learning issues are presented and some experimental results obtained for synthetic realistic data, are reported.
Uninorm Based Fuzzy Network for Tree Data Structures
CIARAMELLA, Angelo;PETROSINO, Alfredo
2009
Abstract
The aim of this study is to introduce a fuzzy model to process structured data. A structured organization of information is typically required by symbolic processing. Most connectionist models assume that data are organized in a form of relatively simple structures such as vectors or sequences. In this work, we propose a connectionist model that can directly process labeled trees. The model is based on a new category of logic connectives and logic neurons that use the concept of uninorms. Uninorms are a generalization of t-norms and t-conorms used for aggregating fuzzy sets. Using a back-propagation algorithm we optimize the parameters of the model (relations and membership functions). The learning issues are presented and some experimental results obtained for synthetic realistic data, are reported.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
