In BAT ANN optimization, fmin becomes so less than fitness value that the condition of comparison of these two quantities is not used.

I am optimizing BAT algorithm with Feed Forward Network. I am facing an issue that the fmin becomes very less than fitness values in the first go that the condition of

if fitness<=fmin

is neglected.

Those who use BAT algorithm know that there are fitness values, fitnessnew values and fmin.

My fmin values are becoming very small in the initial iterations that the above given condition becomes useless and is not implemented.

What should I do in this case, as this is not the issue with PSO, it keeps on decreasing MSE when I used the same fitness function that I am using in BAT.