TY - GEN
T1 - Towards an Adaptive Defuzzification
T2 - 16th International Conference on Modeling Decisions for Artificial Intelligence, MDAI 2019
AU - Torra, Vicenç
AU - Garcia-Alfaro, Joaquin
N1 - Publisher Copyright:
© 2019, Springer Nature Switzerland AG.
PY - 2019/1/1
Y1 - 2019/1/1
N2 - Fuzzy systems have been proven to be an effective tool for modeling and control in real applications. Fuzzy control is a well established area that is used in a large number of real systems. Fuzzy rule based systems are defined in terms of rules in which the concepts that define the rules (both in the antecedent and consequent) can be defined in terms of fuzzy sets. In applications, rules are fired and then a set of consequents need to be combined to make a final decision. This final decision is often computed by means of a defuzzification method. In this paper we discuss the defuzzification proces and propose the use of a Choquet integral for this process. In contrast with standard defuzzification methods which are based on mean operators (usually discrete), the Choquet integral permits us to have an output variable with values that have different importances and with interactions among the values themselves. To illustrate the approach, we use a numerical Choquet integral software for continuous functions that we have recently developed. We also position the application of the approach to handle the uncertainty associated to a mission-oriented Cyber-Physical System (CPS).
AB - Fuzzy systems have been proven to be an effective tool for modeling and control in real applications. Fuzzy control is a well established area that is used in a large number of real systems. Fuzzy rule based systems are defined in terms of rules in which the concepts that define the rules (both in the antecedent and consequent) can be defined in terms of fuzzy sets. In applications, rules are fired and then a set of consequents need to be combined to make a final decision. This final decision is often computed by means of a defuzzification method. In this paper we discuss the defuzzification proces and propose the use of a Choquet integral for this process. In contrast with standard defuzzification methods which are based on mean operators (usually discrete), the Choquet integral permits us to have an output variable with values that have different importances and with interactions among the values themselves. To illustrate the approach, we use a numerical Choquet integral software for continuous functions that we have recently developed. We also position the application of the approach to handle the uncertainty associated to a mission-oriented Cyber-Physical System (CPS).
UR - https://www.scopus.com/pages/publications/85072828047
U2 - 10.1007/978-3-030-26773-5_11
DO - 10.1007/978-3-030-26773-5_11
M3 - Conference contribution
AN - SCOPUS:85072828047
SN - 9783030267728
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 113
EP - 125
BT - Modeling Decisions for Artificial Intelligence - 16th International Conference, MDAI 2019, Proceedings
A2 - Torra, Vicenç
A2 - Narukawa, Yasuo
A2 - Pasi, Gabriella
A2 - Viviani, Marco
PB - Springer Verlag
Y2 - 4 September 2019 through 6 September 2019
ER -