I've checked the PSO-ANFIS can perform the supervised learning as regression and binary classification for multi-input and single-output system (MISO) very well.

However, I have no idea how to expand the PSO-ANFIS(or PSO-RBFN) to MIMO system for multi-decision making. In my opinion, multiple PSO-ANFIS models for each output can be developed, but the computational cost for training would be much larger. Also, I've considered applying Multi-objective PSO (MOPSO), but I have not tried yet.

Can someone recommend any previous research or related web-pages? Thanks.

(Appendix 1) Keywards:

* PSO: Particle swarm optimization

* ANFIS: Adaptive network fuzzy inference system

* RBFN: Radial basis function network

* LSE : Least square estimation

(Appendix 2) The PSO-ANFIS consists of my custom code based on MATLAB.

- PSO for training centers and variations of the Gaussian membership function as known as antecedent parameters.

- Initial values of the antecedent parameters and the number of fuzzy membership functions are determined by Subtractive clustering of MATLAB.

- LSE for training premise parameters.

- Actually, the structure of my PSO-ANFIS is similar to the PSO-RBFN as universal approximator.