Am currently working on a project for simulating fractal antenna

[anug]

am currently working on a project for simulating fractal antenna behaviour using anfis..i got 5 parameters of antenna as the inputs and its return loss as its single output. i got such 150 sets of known inputs and outputs of which i divide it into partly checking sets and training sets..
... i have used genfis2 to create fis structure and training it using anfis
hw does the radii, (range of influence) affect training..
how much should it be (range)
hw many number of epochs before data starts overfitting
i plot the return loss and am getting proper graphs for training data the way it should be but graphs for checking sets it differs a lot from original..hw do i go about it

 

[NateD]

The size of the radii will affect the ability of anfis to accurately model the data. Generally, larger radii can lead to overfitting, while smaller radii may not capture enough of the data. In terms of the number of epochs, it is difficult to determine when the data starts overfitting. However, you can monitor the error rate and use a stopping criterion, such as when the error rate no longer decreases with additional epochs. If the graphs for your checking sets differ significantly from the original, it could be due to overfitting or inadequate training. You may need to experiment with different parameters, such as the number of epochs, the learning rate, or the size of the radii, to see if the results improve.