Adaptive neuro-fuzzy inference system (ANFIS) is a combination of ANN and fuzzy systems
and provides the benefits of two models. But ANFIS in MATLAB is based on Takagi–Sugeno method which produce some limitations, i.e. it can model only one output. Is it possible to develop a mamdani based ANFIS in MATLAB?
Mamdani based ANFIS (M-ANFIS) is applicable and has been of dynamic studying interest so far. It inherits further advantages over T-S FIS methodology. To this aim, you have to constitute a representation to be compliant with M-ANFIS. For instance, Mamdani ANFIS architecture consists of five layers, output of each layer has its corresponding structure. For further information you could refer to the following.
Chai, Y., Jia, L., & Zhang, Z. (2009). Mamdani Model based Adaptive Neural Fuzzy Inference System and its Application. International Journal of Computational Intelligence, 5(1).
I do hope that the answer is constructive and informative enough!
Multiple ANFIS (One ANFIS for each Output) or Co-active NFIS (CANFIS) as suggested by Jang, the original author is NOT available in MATLAB. Unless you want to insert expert-knowledge about the problem domain as Fuzzy Rules it's better to go for neural network first. Try newwff or patternnet in MATLAB.
I am working on ANFIS modeling for a system and MATLAB currently lets me use Sugeno based FIS. I also want to test this on a Mamdani based FIS to check if my results can be improved.
You can find useful information about ANFIS by studying MATLAB software help
- Badges
- Science topic