10-7 模?複?度?決?

[english][all]

pPĤ@`HfuXdҩҭzAXhƷ|sơ]X{Vmơ^wͫܤjvTApUҭzG

]ڭ̻ݭn@ӨtΤƪkӨMwXhơAL{κ٬ҫ]Model Selection^ASOObΩtŲO]System Identification^άO˦{]Pattern Recognition^ɡCb`Aڭ̱Nyz@ӾAΩhXҫkAk٬ud@ժkv]Leave-one-out Test^AiΩ@몺jkDC

]ڭ̦@Ӹƶ $D$A]tF $n$ ӿJXG $$ D=\{ (x_1, y_1), (x_2, y_2), ..., (x_n, y_n) \} $$

ڭ̥iHwq@Ӽҫ $f()$ 󦹸ƶ $D$ RMSE (root-mean-squared error): $$ rmse(f, D) = \sum_{i=1}^n |y_i-f(x_i)|^2 $$

~A polyfit OҦ^ǪhiHܦpUG $$ f = polyfit(D, r)$$

䤤 $D$ OΩXƶA$r$ hOhơC

b{@ɤAƶ $D$ ƵӼƬOA]ڭ̥ƶC@Ƶ[Ρ]ΩҫظmHμҫҡ^AHoƶ̤jĤOC@ӫD`ıkAOä@ơAMϥΩҦLƨӫإ߼ҫAAΤeêƨҼҫAðO RMSEC]ڭ̦ $n$ ơAhڭ̷|إ $n$ ӼҫAo $n$ RMSEAbD䥭 RMSEA٬ Leave-one-out RMSE (LOORMSE)C󤣦PcѼơ]Ҧphơ^Aڭ̥iHإߤPҫAíp LOORMSEA̫ڭ̿ҫcѼơA|ϱo LOORMSE ̤pCSOݭn`NOGLOORMSE uΩܼҫcѼơ]Ҧphơ^ATwcѼƫAڭٻݨϥγ̤pkӨMw̨ΰѼƭȡC

FMaud@ժkvAڭ̥iHwq RMSEG

ЯSO`NA$t$ M $v$ Oh $r$ ơ]άO@ӨAOjkҫתơ^C]ڭ̥iHeX $t$ $v$ $r$ ϧΡAÿX@ $r$ Ȩϱo validating RMSE ̤pȡC]@iѦҥDC^


MATLAB{]pGig