-- card: 8153 from stack: in
-- bmap block id: 8221
-- flags: 0000
-- background id: 2595
-- name: 


-- part 1 (button)
-- low flags: 00
-- high flags: A003
-- rect: left=368 top=133 right=155 bottom=468
-- title width / last selected line: 0
-- icon id / first selected line: 0 / 0
-- text alignment: 1
-- font id: 0
-- text size: 12
-- style flags: 0
-- line height: 16
-- part name: predict err
----- HyperTalk script -----
on mouseUp
  open "bin:strange attractors:Ikeda map:Ikeda.predict.err"
end mouseUp



-- part contents for background part 22
----- text -----
How good is an approximate  ƒ ?
•  We are only interested in the performance of  ƒ  on the attractor itself.   Define the  predictor error  to be




•  For a  D-dimensional attractor,  the typical separation between data points is                         .  Thus a predictor scheme is of  order       if                           is found.
e.g.  for Ikeda,                    (it should be 4,  why is it not?)