I have a summation objective function (non-linear portfolio optimization) which looks like:
minimize w(i)*w(j)*cv(i,j) for i = 1 to 10 and j = 1 to 10
w is the decision vector
cv is a known 10 by 10 matrix
I have the formulation done for the constraints (a separate .m file for the project constraints) and for the execution of the fmincon (a separate .m file for the lower/upper bounds, initial value, and calling fmincon with the arguments).
I just can't figure out how to do the objective function. I'm used to linear programming in GLPK rather than matlab so I'm not doing so good.
I'm currently got:
ObjectiveFunction.m
function f = obj(w)
cv = [all the constants are in here]
i = 1;
j = 1;
n = 10;
var = 0;
while i <= n
while j<=n
var = var + abs(w(i)*w(j)*cv(i, j));
j = j + 1;
end
i = i + 1;
end
f = var
...but this isn't working.
Any help would be appreciated! Thanks in advance :)
So this is from a class I took a few years ago, but it addresses a very similar problem to your own with respect to use of fminsearch to optimize some values. The problem is essentially that you have a t, y, and you want a continuous exponential function to represent t, y in terms of c1*t.*exp(c2*t). The textbook I lifted the values from is called Numerical Analysis by Timothy Sauer. Unfortunately, I don’t remember the exact problem or chapter, but it’s in there somewhere.
c1 and c2 are found recursively by fminsearch minimizing a residual y - ((c1) * t .* exp((c2) * t)). Try copying and running my code below to get a feel for things:
%% Main code
clear all;
t = [1,2,3,4,5,6,7,8];
y = [8,12.3,15.5,16.8,17.1,15.8,15.2,14];
lambda0=[1 -.5];
lambda=fminunc(#expdecayfun,lambda0, ...
optimset('LargeScale','off','Display','iter','TolX',1.e-6),t,y);
c1=lambda(1);
c2=lambda(2);
fprintf('Using the BFGS method through fminunc, c1 = %e\n',c1);
fprintf('and c2 = %e. Since these values match textbook values for\n', c2);
fprintf('c1 and c2, I will stop here.\n');
%% Index of functions:
% expdecayfun
function res=expdecayfun(lambda,t,y) c1=lambda(1);
c2=lambda(2);
r=y-((c1)*t.*exp((c2)*t));
res=norm(r);
Hope this helps!
Related
I am currently working on an assignment where I need to create two different controllers in Matlab/Simulink for a robotic exoskeleton leg. The idea behind this is to compare both of them and see which controller is better at assisting a human wearing it. I am having a lot of trouble putting specific equations into a Matlab function block to then run in Simulink to get results for an AFO (adaptive frequency oscillator). The link has the equations I'm trying to put in and the following is the code I have so far:
function [pos_AFO, vel_AFO, acc_AFO, offset, omega, phi, ampl, phi1] = LHip(theta, eps, nu, dt, AFO_on)
t = 0;
% syms j
% M = 6;
% j = sym('j', [1 M]);
if t == 0
omega = 3*pi/2;
theta = 0;
phi = pi/2;
ampl = 0;
else
omega = omega*(t-1) + dt*(eps*offset*cos(phi1));
theta = theta*(t-1) + dt*(nu*offset);
phi = phi*(t-1) + dt*(omega + eps*offset*cos(phi*core(t-1)));
phi1 = phi*(t-1) + dt*(omega + eps*offset*cos(phi*core(t-1)));
ampl = ampl*(t-1) + dt*(nu*offset*sin(phi));
offset = theta - theta*(t-1) - sym(ampl*sin(phi), [1 M]);
end
pos_AFO = (theta*(t-1) + symsum(ampl*(t-1)*sin(phi* (t-1))))*AFO_on; %symsum needs input argument for index M and range
vel_AFO = diff(pos_AFO)*AFO_on;
acc_AFO = diff(vel_AFO)*AFO_on;
end
https://www.pastepic.xyz/image/pg4mP
Essentially, I don't know how to do the subscripts, sigma, or the (t+1) function. Any help is appreciated as this is due next week
You are looking to find the result of an adaptive process therefore your algorithm needs to consider time as it progresses. There is no (t-1) operator as such. It is just a mathematical notation telling you that you need to reuse an old value to calculate a new value.
omega_old=0;
theta_old=0;
% initialize the rest of your variables
for [t=1:N]
omega[t] = omega_old + % here is the rest of your omega calculation
theta[t] = theta_old + % ...
% more code .....
% remember your old values for next iteration
omega_old = omega[t];
theta_old = theta[t];
end
I think you forgot to apply the modulo operation to phi judging by the original formula you linked. As a general rule, design your code in small pieces, make sure the output of each piece makes sense and then combine all pieces and make sure the overall result is correct.
Finding m and c for an equation y = mx + c, with the help of math and plots.
y is data_model_1, x is time.
Avoid other MATLAB functions like fitlm as it defeats the purpose.
I am having trouble finding the constants m and c. I am trying to find both m and c by limiting them to a range (based on smart guess) and I need to deduce the m and c values based on the mean error range. The point where mean error range is closest to 0 should be my m and c values.
load(file)
figure
plot(time,data_model_1,'bo')
hold on
for a = 0.11:0.01:0.13
c = -13:0.1:-10
data_a = a * time + c ;
plot(time,data_a,'r');
end
figure
hold on
for a = 0.11:0.01:0.13
c = -13:0.1:-10
data_a = a * time + c ;
mean_range = mean(abs(data_a - data_model_1));
plot(a,mean_range,'b.')
end
A quick & dirty approach
You can quickly get m and c using fminsearch(). In the first example below, the error function is the sum of squared error (SSE). The second example uses the sum of absolute error. The key here is ensuring the error function is convex.
Note that c = Beta(1) and m = Beta(2).
Reproducible example (MATLAB code[1]):
% Generate some example data
N = 50;
X = 2 + 13*random(makedist('Beta',.7,.8),N,1);
Y = 5 + 1.5.*X + randn(N,1);
% Example 1
SSEh =#(Beta) sum((Y - (Beta(1) + (Beta(2).*X))).^2);
Beta0 = [0.5 0.5]; % Initial Guess
[Beta SSE] = fminsearch(SSEh,Beta0)
% Example 2
SAEh =#(Beta) sum(abs(Y-(Beta(1) + Beta(2).*X)));
[Beta SumAbsErr] = fminsearch(SAEh,Beta0)
This is a quick & dirty approach that can work for many applications.
#Wolfie's comment directs you to the analytical approach to solve a system of linear equations with the \ operator or mldivide(). This is the more correct approach (though it will get a similar answer). One caveat is this approach gets the SSE answer.
[1] Tested with MATLAB R2018a
I am trying to convert my quadprog linear quadratic problem over to fmincon so that later I can add nonlinear constraints. I am having difficulty when I compare my solutions using the two methods (for the same problem). The odd thing is that I get very different cost output when I get almost the same x values. Below is a simplified case of my code without constraints.
Here, my objective function is
%objective function
% cost = a + b*x(1) + c*x(1)^2 + d + e*x(2) + e*x(2)^2
param = [1;2;3;4;5;6];
H = [2*param(3) 0; 0 2*param(6)];
f = [param(2); param(5)];
x0 = [0,0];
[x1,fval1] = quadprog(H,f);
[x2,fval2] = fmincon(#(x) funclinear(x,param), x0);
fval1
fval2
%% defining cost objective function
function cost = funclinear(x, param);
cost=(param(1) + param(2)*x(1) + param(3)*(x(1))^2+ param(4) +param(5)*x(2)+param(6)*(x(2))^2);
end
My resulting x1 and x2 are
x1 =[-3.333333333305555e-01;-4.166666666649305e-01];
x2 =[-3.333333299126037e-01;-4.166666593362859e-01];
which makes sense that they are slightly different since they are different solvers.
However my optimized costs are
fval1 =-1.375000000000000e+00;
fval2 =3.625000000000001e+00;
Does this mean that my objective function is different than my H and f? Any help would be appreciated.
In the quadprog formulation, the constant terms a and d are not considered.
param(1)+param(4) = 1 + 4 = 5
The difference of your results is also 5.
I have the following problem:
max CEQ(w) s.t. w in (0,1) and I don't know anything about CEQ(w) except that is given by a fixed point equation of the form CEQ(w) = F(CEQ(w)). If I fix a w, I can solve the fixed point equation using the fzero function and obtain a value for CEQ. If I choose a different w, I get another value for CEQ. Thus, I could loop over all possible values of w and then choose the one that gives the highest CEQ. This seems bad programming though and I was wondering whether I can do this more efficient in MATLAB: I want to model the solution to my fixed point equation as a function of w but I don't know how to implement it.
To be more precise, here is a sample code:
clear all
clc
NoDraws = 1000000;
T_hat = 12;
mu = 0.0058;
variance = 0.0017;
rf = 0.0036;
sim_returns(:,T_hat/12) = T_hat*mu + sqrt(T_hat*variance)*randn(NoDraws,1);
A = 5;
kappa=1;
l=0;
theta = 1 - l*(kappa^(1-A) - 1) *(kappa>1);
CEQ_DA_0 = 1.1;
CEQ_opt = -1000;
w_opt = 0;
W_T = #(w) (1-w)*exp(rf*T_hat) + w*exp(rf*T_hat + sim_returns(:,T_hat/12));
for w=0.01:0.01:0.99
W=W_T(w);
fp = #(CEQ) theta*CEQ^(1-A)/(1-A) - mean( W.^(1-A)/(1-A)) + l*mean( ((kappa*CEQ)^(1-A)/(1-A) - W.^(1-A)/(1-A)) .* (W < kappa*CEQ));
CEQ_DA = fzero(fp,CEQ_DA_0);
if CEQ_DA > CEQ_opt
CEQ_opt = CEQ_DA;
w_opt = w;
end
end
That is, in the loop, I fix a w, solve the fixed point equation and store the value for CEQ. If some other w gives a bigger value for CEQ, the current optimal w gets replaced by that new w. what I would like to have (instead of the loop part) is something like this:
fp = #(CEQ,w) theta*CEQ^(1-A)/(1-A) - mean( W_T(w).^(1-A)/(1-A)) + l*mean( ((kappa*CEQ)^(1-A)/(1-A) - W_T(w).^(1-A)/(1-A)) .* (W_T(w) < kappa*CEQ));
CEQ_DA = #(w) fzero(fp,CEQ_DA_0);
[w_opt, fval]=fminbnd(CEQ_DA,0,1);
Your proposed solution is very close. In words, you're defining fp as a function of two arguments, and would like CEQ_DA to be a function of w, which solves fp for CEQ, with that given w. The only issue is that fzero doesn't know which parameter of fp to solve over, because it can't match anonymous function parameters and fp parameters by name.
The answer is yet one more anonymous function inside the fzero, to turn fp(CEP,w) into fp_w(CEP), which will be solveable for CEQ
CEQ_DA = #(w) fzero(#(CEQ) fp(CEQ, w),CEQ_DA_0);
I've been trying to use MATLAB to solve equations like this:
B = alpha*Y0*sqrt(epsilon)/(pi*ln(b/a)*sqrt(epsilon_t))*integral from
0 to pi of
(2*sinint(k0*sqrt(epsilon*(a^2+b^2-2abcos(theta))-sinint(2*k0*sqrt(epsilon)*a*sin(theta/2))-sinint(2*k0*sqrt(epsilon)*b*sin(theta/2)))
with regard to theta
Where epsilon is the unknown.
I know how to symbolically solve equations with unknown embedded in an integral by using int() and solve(), but using the symbolic integrator int() takes too long for equations this complicated. When I try to use quad(), quadl() and quadgk(), I have trouble dealing with how the unknown is embedded in the integral.
This sort of thing gets complicated real fast. Although it is possible to do it all in a single inline equation, I would advise you to split it up into multiple nested functions, if only for readability.
The best example of why readability is important: you have a bracketing problem in the eqution you posted; there's not enough closing brackets, so I can't be entirely sure what the equation looks like in mathematical notation :)
Anyway, here's one way to do it with the version I --think-- you meant:
function test
% some random values for testing
Y0 = rand;
b = rand;
a = rand;
k0 = rand;
alpha = rand;
epsilon_t = rand;
% D is your B
D = -0.015;
% define SIMPLE anonymous function
Bb = #(ep) F(ep).*main_integral(ep) - D;
% aaaand...solve it!
sol = fsolve(Bb, 1)
% The anonymous function above is only simple, because of these:
% the main integral
function val = main_integral(epsilon)
% we need for loop through epsilon, due to how quad(gk) solves things
val = zeros(size(epsilon));
for ii = 1:numel(epsilon)
ep = epsilon(ii);
% NOTE how the sinint's all have a different function as argument:
val(ii) = quadgk(#(th)...
2*sinint(A(ep,th)) - sinint(B(ep,th)) - sinint(C(ep,th)), ...
0, pi);
end
end
% factor in front of integral
function f = F(epsilon)
f = alpha*Y0*sqrt(epsilon)./(pi*log(b/a)*sqrt(epsilon_t)); end
% first sinint argument
function val = A(epsilon, theta)
val = k0*sqrt(epsilon*(a^2+b^2-2*a*b*cos(theta))); end
% second sinint argument
function val = B(epsilon, theta)
val = 2*k0*sqrt(epsilon)*a*sin(theta/2); end
% third sinint argument
function val = C(epsilon, theta)
val = 2*k0*sqrt(epsilon)*b*sin(theta/2); end
end
The solution above will still be quite slow, but I think that's pretty normal for integrals this complicated.
I don't think implementing your own sinint will help much, as most of the speed loss is due to the for loops with non-builtin functions...If it's speed you want, I'd go for a MEX implementation with your own Gauss-Kronrod adaptive quadrature routine.