todo.txt

 <normalhuman> why not just p(n)-p(n-1) ?
[20:24:46] <myke> because dividing it by p(n-1) normalizes the difference
[20:25:17] <myke> 10 to 12 is 2% change
[20:25:48] <myke>  in this case p(n) - p(n-1) is 0.02
[20:25:56] <myke> in below case it is 0.2
[20:25:09] <myke> 1.0 to 1.02 is 2% change

<@myke> ( price(n) - price(n-1) ) / price(n-1)          <--------- price diff (proper - normalized difference)

todo:

check train init weights /random weights / not weighted
jit randomly
fann_create_shortcut/sparse