I have this cubic spline function:
s(x) = 4 + k1*x + 2x^2 - (1/6)*x^3 for x in [0,1]
s(x) = 1 - (4/3)*(x-1) + k2*(x-1)^2 - (1/6) * (x-1)^3 for x in [1,2]
s(x) = 1 + k3*(x-2) + (x-2)^2 - (1/6) * (x-2)^3 for x in [2,3]
I would like to implement a simple function, that giving this function, it determines the coefficients k1, k2 and k3, but I couldn't do that..
Has anyone an idea?
Denoting the 3 functions as s0(x), s1(x) and s2(x), you can expand s1(x) and s2(x) and collect the coefficients for x^3, x^2, x and constant terms. Then, you get
s1(x)= 4 +k1*x+2*x^2-(1/6)*x^3 s2(x)=
(5/2+k2)+(-11/6-2k2)*x+(k2+1/2)*x^2-(1/6)*x^3 s3(x)= ..... (left our
for your own practice)
For these 3 functions to be all cubic polynomials, they should be identical. So, you can have
4 = 5/2 + k2
k1 = -11/6 - 2k2
2 = k2+1/2
from which you can get k1=-29/6, k2=3/2.
Compare the equations s0(x) and s2(x), you can also get k3 = 7/6.
Related
I have an equation: y=ax^3 + bx^2 + cx + d and the list of values x = 1, 2, 3, 4 when y = 3, 4, 3, -6 respectively. In Octave, I want to:
(a) Set up a system of four equations involving a, b, c and d. For example, substituting (x, y) = (1,3) into the polynomial gives the equation 3 = a + b + c + d.
(b) Solve the system in (a).
I've been trying to find how to do this for three hours and found nothing. Any help would be appreciated
Thanks.
pstscrpt - I have to do everything in Octave, even though I could find it by hand
Written without any ; at end of assignements so you can see what is going on.
You problem is basically a linear system in the variables [a,b,c,d]'=z
So you need to build a system A*z=y, where A is a matrix 4x4, y and z are column vector size 4
x=[1,2,3,4]'
y=[3,4,3,-6]'
A=zeros(4,4)
for i=1:4
A(i,:)= [ x(i)^3, x(i)^2, x(i), 1]
endfor
z=A\y
the outcome will be
z =
-1.00000
5.00000
-7.00000
6.00000
In Matlab: start by just substituting the different values of x and y you wrote in the expression a*x^3 + b*x^2 + c*x + d = y as:
syms a b c d
eqn1 = a*1^3 + b*1^2 + c*1^1 +d == 3 ;
eqn2 = a*2^3 + b*2^2 + c*2^1 +d == 4 ;
eqn3 = a*3^3 + b*3^2 + c*3^1 +d == 3 ;
eqn4 = a*4^3 + b*4^2 + c*4^1 +d == -6 ;
Then Use equationsToMatrix to convert the equations into the form AX = B. The second input to equationsToMatrix specifies the independent variables in the equations.:
[A,B] = equationsToMatrix([eqn1, eqn2, eqn3, eqn4], [a, b, c,d ])
and the solution for a,b,c,d is:
X = linsolve(A,B)
you can also use if you want
sol = solve([eqn1, eqn2, eqn3, eqn4], [a, b, c,d ])
Good afternoon!
First things first, I looked for similar questions for a while, but (probably because of my inexperience) I've found nothing similar to what I'm going to ask.
I'm using matlab for the first time to solve this kind of problems, so I'm not sure of what to do. A brief explenation:
I'm doing a project for my Optimal Control course: I have to replicate the results of a paper about employment, and I'm stuck with the plots. I have the following data:
five variable functions (U(t), T(t), R(t), V1(t) and V2(t))
four control functions(u1(t), u2(t), u3(t), u4(t))
constraints on the control variables (each u must be between 0 and 1)
initial values for U, T, R, V1 and V2 (in t=0, in particular V1 and V2 are constant over time)
final values for the λ coefficients in the hamiltonian
(note: for the controls, I've already found the optimal expression, which is in this form: ui = min{1, max{0,"expression"}}. If needed, I can give also the four expressions, neglected to
synthesize a little)
Under professor's suggestions, I've tried to use fmincon, that theoretically should give me directly the information that I need to plot some result using only the cost function of the problem. But in this case I have some issues involving time in the calculations. Below, the code that I used for fmincon:
syms u
%note: u(5) corresponds to U(t), but this is the only way I've found to get
%a result, the other u(i) are in ascending order (u(1) = u1 and so on...)
g = #(u) 30*u(5) + (20/2)*(u(1))^2 + (20/2)*(u(2))^2 + (10/2)*(u(3))^2 + (40/2)*(u(4))^2;
%initial guesses
u0 = [0 0 0 0 100000]; %
A = [];
b = [];
Aeq = [];
beq = [];
lb = 0.0 * ones(1,2,3,4);
ub = 1.0 * ones(1,2,3,4);
[x,fval,output,lambda] = fmincon(g, u0, A, b, Aeq, beq, lb, ub);
Whit this code, i get (obviously) only one value for each variable as result, and since I've not found any method to involve time, as I said before, I start looking for other solving strategies.
I found that ode45 is a differential equation solver that has the "time iteration" already included in the algorithm, so I tried to write the code to get it work with my problem.
I took all the equations from the paper and put them in a vector as shown in the mathworks examples, and this is my matlab file:
syms u1(t) u2(t) u3(t) u4(t)
syms U(t) T(t) R(t) V1(t) V2(t)
syms lambda_u lambda_t lambda_r lambda_v1 lambda_v2
%all the parameters provided by the paper
delta = 500;
alpha1 = 0.004;
alpha2 = 0.005;
alpha3 = 0.006;
gamma1 = 0.001;
gamma2 = 0.002;
phi1 = 0.22;
phi2 = 0.20;
delta1 = 0.09;
delta2 = 0.05;
k1 = 0.000003;
k2 = 0.000002;
k3 = 0.0000045;
%these two variable are set constant
V1 = 200;
V2 = 100;
%weight values for the cost function (only A1 is used in this case, but I left them all since the unused ones are irrelevant)
A1 = 30;
A2 = 20;
A3 = 20;
A4 = 10;
A5 = 40;
%ordering the unknowns in an array
x = [U T R u1 u2 u3 u4];
%initial conditions, ordered as the x vector (for the ui are guesses)
y0 = [100000 2000 1000 0 0 0 0];
%system set up
f = #(t,x) [delta - (1 + x(4))*k1*x(1)*V1 - (1 + x(5))*k2*x(1)*V2 - alpha1*x(1) + gamma1*x(2) + gamma2*x(3);...
(1 + x(4))*k1*x(1)*V1 - k3*x(2)*V2 - alpha2*x(2) - gamma1*x(2);...
(1 + x(5))*k2*x(1)*V2 - alpha3*x(3) - gamma2*x(3) + k3*x(2)*V2;...
alpha2*x(2) + gamma1*x(2) + (1 + x(6))*phi1*x(1) + k3*x(2)*V2 - delta1*V1;...
alpha3*x(3) + gamma2*x(3) + (1 + x(7))*phi2*x(1) - delta2*V2;...
-A1 + (1 + x(4))*k1*V1*(lambda_u - lambda_t) + (1 + x(5))*k2*V2*(lambda_u - lambda_r) + lambda_u*alpha1 - lambda_v1*(1 + x(6))*phi1 - lambda_v2*(1 + x(7))*phi2;...
-lambda_u*gamma1 + (alpha2 + gamma1)*(lambda_t - lambda_v1) + k3*V2*(lambda_t - lambda_r - lambda_v1);...
-lambda_u*gamma2 + (alpha3 + gamma2)*(lambda_r - lambda_v2);...
(1 + x(4))*k1*x(1)*(lambda_u - lambda_t) + lambda_v1*delta1;...
(1 + x(5))*k2*x(1)*(lambda_u -lambda_r) + k3*x(2)*(lambda_t - lambda_r - lambda_v1) + lambda_v2*delta2];
%using ode45 to solve over the chosen time interval
[t,xa] = ode45(f,[0 10],y0);
With this code, I get the following error:
Error using odearguments (line 95)
#(T,X)[DELTA-(1+X(4))*K1*X(1)*V1-(1+X(5))*K2*X(1)*V2-ALPHA1*X(1)+GAMMA1*X(2)+GAMMA2*X(3);(1+X(4))*K1*X(1)*V1-K3*X(2)*V2-ALPHA2*X(2)-GAMMA1*X(2);(1+X(5))*K2*X(1)*V2-ALPHA3*X(3)-GAMMA2*X(3)+K3*X(2)*V2;ALPHA2*X(2)+GAMMA1*X(2)+(1+X(6))*PHI1*X(1)+K3*X(2)*V2-DELTA1*V1;ALPHA3*X(3)+GAMMA2*X(3)+(1+X(7))*PHI2*X(1)-DELTA2*V2;-A1+(1+X(4))*K1*V1*(LAMBDA_U-LAMBDA_T)+(1+X(5))*K2*V2*(LAMBDA_U-LAMBDA_R)+LAMBDA_U*ALPHA1-LAMBDA_V1*(1+X(6))*PHI1-LAMBDA_V2*(1+X(7))*PHI2;-LAMBDA_U*GAMMA1+(ALPHA2+GAMMA1)*(LAMBDA_T-LAMBDA_V1)+K3*V2*(LAMBDA_T-LAMBDA_R-LAMBDA_V1);-LAMBDA_U*GAMMA2+(ALPHA3+GAMMA2)*(LAMBDA_R-LAMBDA_V2);(1+X(4))*K1*X(1)*(LAMBDA_U-LAMBDA_T)+LAMBDA_V1*DELTA1;(1+X(5))*K2*X(1)*(LAMBDA_U-LAMBDA_R)+K3*X(2)*(LAMBDA_T-LAMBDA_R-LAMBDA_V1)+LAMBDA_V2*DELTA2]
returns a vector of length 10, but the length of initial conditions vector is 7. The vector returned by
#(T,X)[DELTA-(1+X(4))*K1*X(1)*V1-(1+X(5))*K2*X(1)*V2-ALPHA1*X(1)+GAMMA1*X(2)+GAMMA2*X(3);(1+X(4))*K1*X(1)*V1-K3*X(2)*V2-ALPHA2*X(2)-GAMMA1*X(2);(1+X(5))*K2*X(1)*V2-ALPHA3*X(3)-GAMMA2*X(3)+K3*X(2)*V2;ALPHA2*X(2)+GAMMA1*X(2)+(1+X(6))*PHI1*X(1)+K3*X(2)*V2-DELTA1*V1;ALPHA3*X(3)+GAMMA2*X(3)+(1+X(7))*PHI2*X(1)-DELTA2*V2;-A1+(1+X(4))*K1*V1*(LAMBDA_U-LAMBDA_T)+(1+X(5))*K2*V2*(LAMBDA_U-LAMBDA_R)+LAMBDA_U*ALPHA1-LAMBDA_V1*(1+X(6))*PHI1-LAMBDA_V2*(1+X(7))*PHI2;-LAMBDA_U*GAMMA1+(ALPHA2+GAMMA1)*(LAMBDA_T-LAMBDA_V1)+K3*V2*(LAMBDA_T-LAMBDA_R-LAMBDA_V1);-LAMBDA_U*GAMMA2+(ALPHA3+GAMMA2)*(LAMBDA_R-LAMBDA_V2);(1+X(4))*K1*X(1)*(LAMBDA_U-LAMBDA_T)+LAMBDA_V1*DELTA1;(1+X(5))*K2*X(1)*(LAMBDA_U-LAMBDA_R)+K3*X(2)*(LAMBDA_T-LAMBDA_R-LAMBDA_V1)+LAMBDA_V2*DELTA2]
and the initial conditions vector must have the same number of elements.
Error in ode45 (line 115)
odearguments(FcnHandlesUsed, solver_name, ode, tspan, y0, options, varargin);
Error in test (line 62)
[t,xa] = ode45(f,[0 10],y0);
For which I can't find a solution, since I have used all the initial values given in the paper. The only values that I have left are the final values for the lambda coefficients, since they are final values, and I am not sure if they can be used.
In this case, I can't also understand where I should put the bounds on the control variable.
For completeness, I will provide also the link to the paper in question:
https://www.ripublication.com/ijss17/ijssv12n3_13.pdf
Can you help me figure out what I can do to solve my problems?
P.S: I know this is a pretty bad code, but I'm basing on the basics tutorials on mathworks; for sure this should need to be refactored and ordered in various file (one for the cost function and one for the constraints for example) but firstly I would like to understand where the problem is and then I will put all in a pretty form.
Thank you so much!
Generally you confused something with Vectors. In initial conditions you declared 7 values:
%initial conditions, ordered as the x vector (for the ui are guesses)
y0 = [100000 2000 1000 0 0 0 0];
But you declared 10 ODE's:
%system set up
f = #(t,x) [delta - (1 + x(4))*k1*x(1)*V1 - (1 + x(5))*k2*x(1)*V2 - alpha1*x(1) + gamma1*x(2) + gamma2*x(3);...
(1 + x(4))*k1*x(1)*V1 - k3*x(2)*V2 - alpha2*x(2) - gamma1*x(2);...
(1 + x(5))*k2*x(1)*V2 - alpha3*x(3) - gamma2*x(3) + k3*x(2)*V2;...
alpha2*x(2) + gamma1*x(2) + (1 + x(6))*phi1*x(1) + k3*x(2)*V2 - delta1*V1;...
alpha3*x(3) + gamma2*x(3) + (1 + x(7))*phi2*x(1) - delta2*V2;...
-A1 + (1 + x(4))*k1*V1*(lambda_u - lambda_t) + (1 + x(5))*k2*V2*(lambda_u - lambda_r) + lambda_u*alpha1 - lambda_v1*(1 + x(6))*phi1 - lambda_v2*(1 + x(7))*phi2;...
-lambda_u*gamma1 + (alpha2 + gamma1)*(lambda_t - lambda_v1) + k3*V2*(lambda_t - lambda_r - lambda_v1);...
-lambda_u*gamma2 + (alpha3 + gamma2)*(lambda_r - lambda_v2);...
(1 + x(4))*k1*x(1)*(lambda_u - lambda_t) + lambda_v1*delta1;...
(1 + x(5))*k2*x(1)*(lambda_u -lambda_r) + k3*x(2)*(lambda_t - lambda_r - lambda_v1) + lambda_v2*delta2];
Every line in above code is recognized as one ODE.
But that's not all. The second problem is with your construction. You mixed symbolic math (lambda declared as syms) with numerical solving, which will be tricky. I'm not familiar with the exact scientific problem you are trying to solve, but if you can't avoid symbolic math, maybe you should try dsolve from Symbolic Math Toolbox?
How explicitly does Matlab solve the rightmost equation in c1, specifically ((x-1)\y)?
I am well aware what happens when you use a matrix, e.g. A\b (where A is a matrix and b is a column vector), but what happens when you use backslash on two vectors with equal rows?
The problem:
x = (1:3)';
y = ones(3,1);
c1 = ((x-1)\y) % why does this =0.6?
You're doing [0;1;2]\[1;1;1]. Essentially x=0.6 is the least-squares solution to
[0;1;2]*x=[1;1;1]
The case you have from the documentation is the following:
If A is a rectangular m-by-n matrix with m ~= n, and B is a matrix with m rows, then A\B returns a least-squares solution to the system of equations A*x= B.
(Specifically, you have m=3, n=1).
A = (1:3).' - 1; % = [0;1;2]
B = ones(3,1); % = [1;1;1]
x = A\B; % = 0.6
Algebraically, it's easy to see this is the solution to the least-squares minimisation
% Calculate least squares
leastSquares = sum( ((A*x) - B).^2 )
= sum( ([0;1;2]*x - [1;1;1]).^2 )
= sum( [-1; x-1; 2x-1].^2 )
= 1 + (x-1)^2 + (2x-1)^2
= 1 + x^2 - 2*x + 1 + 4*x^2 - 4*x + 1
= 5*x^2 - 6*x + 3
% Minimum least squares -> derivative = 0
d(leastSquares)/dx = 10*x - 6 = 0
10*x = 6
x = 0.6
I have no doubt that MATLAB uses a more sophisticated algorithm to come to the same conclusion, but this lays out the mathematics in a fairly plain way.
You can see experimentally that there is no better solution by testing the following for various values of x... 0.6 gives the smallest error:
sum( ([0;1;2]*x - [1;1;1]).^2 )
I am asked to use the least squares method to fit the parameters α and β in y = α*exp(-β*x),
given the points:
x = [1 2 3 4 5 6 7]
y = [9 6 4 2 4 6 9]
I am having trouble determining what my matrix should look like. I know I should take the natural logarithm of both sides of the function in order to get rid of the exponential, and also obtain the natural logarithm of the y-values, which are:
ln_y = [2.19 1.79 1.39 0.69 1.39 1.79 2.19]
However what should my matrix look like, because what I am left with is
ln(y) = ln(α) - β*x?
So the -β column consists of ones and the x column will be my x values, but what should the α column contain?
This is what I assume I should get:
A = [1 1 1 1 1 1 1; 1 2 3 4 5 6 7]
Am I thinking correctly?
The first thing we can do is to take the natural logarithm ln (log in Matlab)) on both sides of the equation:
y = α * e^(-β * x)
becomes:
ln(y) = ln(α * e^(-β * x))
// Law of logarithms
ln(x * y) = ln(x) + ln(y)
// thus:
ln(y) = ln(α) + ln(e^(-β * x))
Simplifying:
ln(y) = -β * x + ln(α)
Now we have ln(y) as a linear function of x and the problem reduces to finding the linear regression in the least square sense. Let's define lny = log(y), and A = ln(α) and we can rewrite the problem as
lny = -β * x + A
Where
x = [1 2 3 4 5 6 7]
lny = [2.19 1.79 1.39 0.69 1.39 1.79 2.19]
For each x_i in x we can evaluate lny as follows (rewritten in ascending power of x):
lny(x1) = A - β * x1
lny(x2) = A - β * x2
...
lny(xn) = A - β * xn
In matrix form
LNY = X * [A β]'
Or,
X * [A β]' = LNY
// let Coefs = [A β]'
Coefs = X^-1 * LNY
In Matlab
x = [1 2 3 4 5 6 7];
y = [9 6 4 2 4 6 9];
lny = log(y);
X = [ones(length(y), 1), -x']; % design matrix
coefs = X\lny'
% A = coefs(1) and β = coefs(2)
% ln(α) = A thus α = exp(A)
alpha = exp(coefs(1));
beta = coefs(2)
You almost had it. The second row should be -x.
x = [1 2 3 4 5 6 7]
y = [9 6 4 2 4 6 9]
logy = log(y)
n = length(x);
A = [ones(1,n); -x]
c = logy/A; %Solve for coefficients
alpha = exp(c(1))
beta = c(2);
In this example, deriving the least squares estimator is a good idea. The other answers take this approach.
There is quick and dirty approach that is flexible and handy.
Just to it numerically. You can use fminsearch to get the job done.
% MATLAB R2019a
x = [1 2 3 4 5 6 7];
y = [9 6 4 2 4 6 9];
% Create anonymous function (your supposed model to fit)
fh =#(params) params(1).*exp(-params(2).*x);
% Create anonymous function for Least Squares Error with single input
SSEh =#(params) sum((fh(params)-y).^2); % Sum of Squared Error (SSE)
p0 = [1 0.5]; % Initial guess for [alpha, beta]
[p, SSE] = fminsearch(SSEh,p0);
alpha = p(1); % 5.7143
beta = p(2); % 1.2366e-08 (AKA zero)
It is always a good idea to plot the results as a sanity check (I screw up often and this saves me time and time again).
yhath=#(params,xval) params(1).*exp(-params(2).*xval);
Xrng = min(x)-1:.2:max(x)+1;
figure, hold on, box on
plot(Xrng,p(1).*exp(-p(2).*Xrng),'r--','DisplayName','Fit')
plot(x,y,'ks','DisplayName','Data')
legend('show')
A Note on Nonlinearity:
This works fine with linear models due to convexity. If your error function is nonlinear but convex, as Sum of Squared Error (SSE), then this also returns the global optimum.
Note that a non-convex function would require multiple start points to attempt to capture many local optima, then taking the best one would still carry no guarantees of optimality. Adding constraints to the solution would require penalty functions or switching to the constrained solver since fminsearch solves the unconstrained problem (unless you penalize it properly).
Easy to Modify:
It is easy to modify the model and the error function. For example, if you wanted to minimize the sum of the absolute error instead, it is straightforward using abs.
% Create anonymous function for Least Absolute Error with single input
SAEh =#(params) sum(abs(fh(params)-y)); % Sum of Absolute Error
I have a general equation
t=tr+(ts-tr)/(1+(a*h)^n)^(1-1/n)
for (h=0, 1, 2, 3), I have t=2.000, 1.6300, 1.2311, 1.1084. therefor there are 4 equations with 4 unknowns tr, ts, a, n
I used "solve" function in matlab
s=solve('tr+(ts-tr)/(1+(a*0)^n)^(1-1/n)=2','tr+(ts-tr)/(1+(a*1)^n)^(1-1/n)=1.63','tr+(ts-tr)/(1+(a*2)^n)^(1-1/n)=1.2311','tr+(ts-tr)/(1+(a*3)^n)^(1-1/n)=1.1084')
and error is
??? Error using ==> mupadmex
Error in MuPAD command: Singularity [ln];
during evaluation of 'numeric::fsolve'
Error in ==> sym.sym>sym.mupadmexnout at 2018
out = mupadmex(fcn,args{:});
Error in ==> solve at 76
[symvars,R] = mupadmexnout('symobj::solvefull',eqns,vars);
What should I do?
The problem appears with you using the solve function. That only works for simple equations, it is better to use the fsolve function. Due to the fact that I am worried that I am doing an assignment for you, I am only going to show you how to do another example using fsolve.
Suppose that you want to solve
1 = x_1
1 = x_1 + x_2
-1 = x_1 + x_2 + x_3
-1 = x_1 + x_2 + x_3 + x_4
then what you firstly need to do is write these as equations equal 0
0 = x_1 - 1
0 = x_1 + x_2 - 1
0 = x_1 + x_2 + x_3 + 1
0 = x_1 + x_2 + x_3 + x_4 + 1
then you need to write a function that takes in a vector x, the components of x will represent x_1, x_2, x_3 and x_4. The output of the function will also be a vector whose components should the outputs of the Right hand side of the above equations (see the function fun below). This function is going to be called by fSolve for it to provide it with guesses of the correct value of x, until it guess correct. When never actually run this function ourselves. That is why it is below the top function.
Then you create a function handle to this function by fHandle = #fun. You can think of fHandle as another name for fun, when we calculate fHandle([1; 2; 3; 4]) this is the same as calculating fun([1; 2; 3; 4]). After this you make an initial guess of the correct vector x, say we chose xGuess = [1; 1; 1; 1]. Finally we pass fHandle and xGuess to fSolve.
Here is the code
function Solve4Eq4Unknown()
fHandle = #fun;
xGuess = ones(4,1);
xSolution = fsolve(fHandle, xGuess)
end
function y = fun(x)
y = zeros(4,1); % This step is not necessary, but it is effecient
y(1) = x(1) - 1;
y(2) = x(1) + x(2) - 1;
y(3) = x(1) + x(2) + x(3) + 1;
y(4) = x(1) + x(2) + x(3) + x(4) + 1;
end