Interpreting BPNN indicator

[Dexter]

Hi everybody, as we know gpwr posted an indicator based on back-propogation neural network which was applied to moving average. I'm trying to build an EA using this indicator, but the problem was raised on backtesting. When you apply an indicator, the network only uses some amount of bars to train itself. Due to this, on the backtest, the indicator does not get values on my specified dates. For example: when I need to test on H4 2009.06.01 - 2009.11.01, the indicator values are only from 2009.10.27. And the Journal gives me Divide by zero errors. Does someone have solution to this? Helps will be greatly appreciated. Thank you