I'm interested in high-order TSK. Does it really have any advantage as compared with zero or first order TSK, or neuro fuzzy in modelling?
Hi
I found a papers about application of quadratic equation in consequent part of a fuzzy system before. In that paper they compared 2nd order tsk system with the 1st order in their application. Beacuse of quadratic nature of second order tsk we have expected its better performance.
Ref:
Estimating the concentration of optically active constituents of sea water by ts models with quadratic consequents; marco cococcioni, et al. Pattern recognition 2007
Interesting question!
I think higher-order TSK deserve to be used and evaluated more extensively. I see many potentials. There are many reasons why zero-order are popular (they are very low computationally complex to identify) and first order are popular as well (they have nice properties in controlling piece-wise linearized systems). However my intuition suggests me that probably there are situations where second order (and even third order) are very promising. The use of orders higher than three are more difficult to be justified.
You can use it in classification methos. And mamdani cannot be a good way to do that
hi,Dr.GholamAli
I am a reader of postgraduate from Jiangnan university of China. Recently, I am reading your paper published on IEEE Transactions on Fuzzy System named "New Formulation for Representing Higher Order TSK Fuzzy Systems". The idea of the paper is remarkable and I am really inspired by your idea. However, I am caught in some problemms about the paper.
Firstly , I am not very good at programming .So in the simulation speed comparsion part, I cannot get charts like the paper presenting at page 6. So may I have your codes about this part??? I will appreciate it! Secondly, in the ANFIS on high order TSK systems part at page 7, about the step 1) of two-input two-order TSK represented by three first-order TSK systems of producer ,I do not know what the "sensitivity of g" is that? And how can I determine it? There are formulations about sensitivity or something else? Thirdly,in the Example 1 at page 8,why are there a total number of 20 parameters in each systems? I hold the opinion that there are 12 consequent parameters in each systems and 3 premise parameters (namely a ,b and c in MFs) means total 15 not 20 parameters to be found? Lastly, at the Table1: SUmmary of the first example,the Rules and epochs are set by empirical value or artifically?
I am looking forward to your reply. Thanks a lot again!!!!
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Control Theory
- Control Systems