Use the tester to optimize net weights

 

In theory , and i'm no expert in anything , if i build a neural network , the basic feed forward mechanism only , and i let the genetic algorithm of mt5 

work on the weights without any backpropagation i'm still "optimizing" the net right ? I mean , i'm trying to guess the weights , but i'm still optimizing .

(prerequisites being custom signal structures and trade structures and custom criterion , and limitations being small sample (the chart the tester is running on only) and the 10500 something passes)

EDIT : stupid idea , lol 

i'll attach the set file when it shows signs of life ...  😂

 

Well it seems it's doable but in 11 hours maybe back prop could have achieved more .  ☕️

a schematic of what was going on in the signal process :

The tester was tasked with optimizing the gray weights not the blue nodes 

Edit : Final image , looks like a cheesecake 

Files:
Reason: