Differences on Fuzzy evaluation between fis.c and Matlab Fuzzy.

Greetings to all.

I have used the anfis tool to train a fuzzy system. The result was a Tagaki-Sugeno type system with 4 inputs, 1 output, 3 gaussian membership functions on each input and 81 rules.

The resulted *.fis file is now my trained fuzzy system. When i use the evalfis() function to evaluate my testing data everything seems to be ok. So far, so good.

In order to embed this *.fis on my application i used the C standalone fuzzy library of matlab (which is fis.c). The problem is that the results i get from fis.c (for the same data set and same fis) are different from those of Matlab evalfis() and do not seem correct.

This error only occurs with Tagaki-Sugeno fuzzy systems. I tested a simple mamdani type fuzzy and the results were similar. So, my guess is that there is a bug on fis.c with sugeno type systems.

Has anybody else faced this problem?

I am using 2012a.

Thank you Fotis