|
|
EDA365欢迎您登录！
您需要 登录 才可以下载或查看，没有帐号？注册
x
求教大神：8 h j3 h& V. B8 F
function f=nsllh(x)( y X8 L- A) A1 \/ K& V
close=xlsread('HW_3_Part2Data','#3','D3 1257');+ O8 H, X9 r9 m
nvix=xlsread('HW_3_Part2Data','#3','H4:H1257');
: x k7 R* h( ~) i# o9 J; talpha=zeros(1,1);beta=zeros(1,1);theta=zeros(1,1);gamma=zeros(1,1);! ^3 e3 s# H+ Q
x=[alpha,beta,theta,gamma]; `1 ^- s& `% T$ v
r=zeros(1254,1);
3 Q( n% e. l$ O- m- c5 n& p7 f( Zfor i=1:1254
% O! b/ u% o2 E6 V& Q r(i+1,1)=log(close(i+1,1)/close(i,1));
3 T- I. `& d' P+ a+ i6 ^end0 m1 l1 a4 Z7 j5 I6 ^
sigma=var(r);
1 ~+ K3 J8 o6 I5 }0 i& jomega=sigma*(1-alpha*(1+theta^2)-beta);" S9 U$ {+ [$ s3 j' @
npsigma=ones(1254,1);/ s6 s& i! A1 ~' w r+ ~; E% K0 K
npsigma(1,1)=sigma;0 m2 t, a- e) n0 Z$ n
for i=1:1253( T% @* g& E1 i5 T; p& L' T7 W
npsigma(i+1,1)=omega+alpha*(r(i,1)-theta*npsigma(i,1)^0.5)^2+beta*npsigma(i,1)+gamma*nvix(i,1)^2/252;: w5 S& N- r7 ]$ G N& Q
end' y7 X1 e5 m0 N" k
alpha=x(1);beta=x(2);theta=x(3);gamma=x(4);' \2 v: F9 c1 D6 g- I& Q/ F
for i=1:1254
/ W" `; n# s: _$ Y' K- unllh(i,1)=0.5*log(2*3.1415926)+0.5*log(npsigma(i,1))+0.5*(r(i,1)^2/npsigma(i,1));2 C# u4 c- g/ I/ u3 A
end
" {2 Q( U: t& m& W# F6 hf=sum(nllh);
% D8 d l/ B( ]
$ }3 J+ }: S3 ^3 ]) [另外nonlcon的code是：2 s# f0 _3 P* H: q2 ]
function [c,ceq]=mycon(x)
! h+ v' ~( z% t# G$ ?; malpha=zeros(1,1);beta=zeros(1,1);theta=zeros(1,1);
$ R( K3 x$ E, M o, @5 D' A% Lx=[alpha,beta,theta];
3 Z" q1 h& T2 o/ N& V/ h) pc=alpha*(1+theta^2)+beta-1;
/ I, m- \# o2 N9 _' ], ?$ `6 cceq=[];/ ~; b! Z. i" _$ r
3 G4 e k9 L8 B: A7 D4 K1 X
run出来显示：9 n9 [) o P! M9 F& \( P$ f
x0=[0.04;0.5;2;0.07];) _( k5 I" R4 t9 }
Optimization terminated: first-order optimality measure less& n4 v! ]& ]" I" U: N
than options.TolFun and maximum constraint violation is less+ h+ b) `! t, p, Y( x3 w' \
than options.TolCon.
% R1 m: l3 N. `, `, s7 FNo active inequalities.0 A2 D+ p X7 K
2 d; |1 l. l" r% z
x =; Y8 ^ x7 r, _- L; T
8 s) W6 E% [2 T/ d7 T' ^; A6 c0 l0 _ 0.0400
, g5 M# V3 p4 j% s 0.5000
4 `1 ?6 `/ s: u* P% M2 Q, b 2.0000& b4 m" t* x1 \" V6 k' P
0.0700
5 }; W! h% T7 F% k
0 y/ S# z6 I# D4 C) x/ b# q; y
6 f [1 A) O4 J& Q3 l# Zfval =1 O4 F, P/ N! N+ U+ M2 Q2 J
& g; R) {, t6 p) w1 i6 ~8 v-3.6963e+003: ?8 o* V& \3 z2 |* M
! ~& g3 t! _7 t% ?( f' Q5 Q4 W: ^' [- W( I
exitflag = l! P; a0 Y" m: D Z
+ n9 ~5 E. m2 P2 g
1
2 t& V5 p# k$ t7 h
- ~+ g$ K% f+ y. u+ b( R
4 y8 v6 N9 M9 C8 ^5 ~. z6 Aoutput =
, [- i2 v# U" P+ @+ q' J. Q5 X/ x T
iterations: 1
% q" ?/ ^( R& [5 } w4 \4 x funcCount: 10 X, K, P' l( c% x
stepsize: 1& L I4 o- O6 o
algorithm: 'medium-scale: SQP, Quasi-Newton, line-search'" s( ^' T# u0 j3 A# V: Y
firstorderopt: 0
5 D4 {0 S! j2 j8 p cgiterations: []
& y# J& A- c& o! U" F7 R message: [1x144 char]8 S9 z. q+ ]3 x9 |! z
# u9 ^7 L4 T( C
7 T' u! K. t# |. slamda =
; Q- O% S8 U4 Z% y' O
5 p$ P. t9 S4 d4 K6 P/ N lower: [4x1 double]; W5 C- }" _. G# p6 K
upper: [4x1 double]* \3 I9 Z5 ]! S* _" z* R
eqlin: [1x0 double]. `; B+ Q {4 s( r7 e/ ^
eqnonlin: [1x0 double]
4 `1 e3 A' q) o0 X. x3 @# M ineqlin: [1x0 double]4 s- Y; B' X5 K$ G# A
ineqnonlin: 0
% h3 F. z; p H
. m P3 N [0 w* ?. j }# ?为什么会这样呢？fval已经改变了但是x不会变？而且iteration只有1
& K7 Y) H) r( G2 w# ~) I# {) G7 ~2 x/ Q" F6 B! I2 `
|
|
/1 
发表于 2020-8-3 14:12
收藏
淘帖
支持！
反对！
楼主