[Hans Blom, 950914b]
(Bill Powers (950913.1100 MDT))
Below is the code for my test program.
Bill, I have solved THIS problem already in my "prediction" post,
where I showed you that the RMS error would be something like 0.7.
Let me demonstrate this to you in a few words. The central equation
in your program is
x := u + d[i]; { ENVIRONMENT}
Since the model knows both x and u directly, it therefore also knows
d[i], through a simple subtraction: d[i] = x - u. It knows d[i] with
full precision, NOT just an estimate. Therefore my earlier program
applies. It predicts the next noise sample d[i+1] based on the values
of d[i] (known from the subtraction) and d[i-1] (memorized).
Can we think of something new? An unknown "world gain", for instance?
Greetings,
Hans