Can I set the input bounds for fuzzy systems trained by ANFIS?

I am using Matlab to generate and train a fuzzy inference system (FIS).

First, I use genfis2 to generate the initial FIS, and then train it by using anfis.

The xBounds is set to [0, 100] or [0, 1], the inputs are percentage or binary features. The following picture shows the initial FIS as generated by genfis2.

.

.

However, after training this initial FIS, the input bound for the second input changes to [0, 50], since 50 is the largest value for the second feature in the subset of the dataset used for training. The FIS after training by anfis becomes as follows:

.

.

This is a problem because when I use the trained fuzzy system on a test dataset, I get the following warning:

Warning: Some input values are outside of the specified input range.
In evalfis at 76

Is it possible to set the value range for input features in anfis (as done with genfis2 by using xBounds) ??

Any help would be appreciated.

.

[Francisco Ferreira]

Sultan Ibrahim, as your message was posted more than 1 year ago, did you manage to solve this problem? I am facing the same problem. I am using code to train the anfis network and get a pretty good accuracy on the data, but when I evaluate the model on the test data, my accuracy is really bad and i get the same warning as you.