I used ode45 and ode23 for a set of non-stiff differential equations. However, both methods converge to a slightly different solution. How can I find out which one is correct? See attached plot where blue is ode45, red is ode23. The dashed lines are the final values for each solver. Also, ode15s is a bit different (less than 1 m)...
Matlab's ODE solvers are adaptive so one specifies tolerances rather than a step size (see also this answer). Given the code in the PDF linked in the comments, if you specify a smaller value for the relative tolerance, the solutions from ode45 and ode23 will converge after the same amount of time. You can use odeset to set 'RelTol':
...
opts = odeset('RelTol', 1e-12);
[t, oput] = ode23(#(t,y)secondode(t,y,C,g), tspan, IC, opts);
...
Note that I also got rid of the global variables used in the linked code (they're bad and inefficient). You also need to change the function definition for secondode to:
function z = secondode(t, indata, C, g)
...
Related
I have a task of writing an M script that would set up (and finally solve) a set of 19 differential equations and all of the equation coefficients. I am not sure what is the best way to input those equations.
Example of some equations:
Example of coefficients:
I don't think that this would be suited for Simulink. Other examples use a function in the form of #(t,x) where t is time and x the vector of all variables.
There are loads of "examples" online, but they don't seem to be appropriate for such a large set of large equations.
For example this exmple of solving 3 equatons
Even for 3 simple equations as they have solved, the functions are getting messy:
f = #(t,x) [-x(1)+3*x(3);-x(2)+2*x(3);x(1)^2-2*x(3)];
Using this notation and getting to the x(19) and cross-referencing all instances of x would be a mess.
I would like your help, and a simple example of how could I write these equations line by line, maybe using the symbolic toolbox, and then put them in an array that I can then forward to the solver.
As I said, I know there are examples online, but they tackle only the most basic system, and really they don't scale well if you want a clean and easily readable code.
I would like to have a similar to Wolfram alpha, where you type variable names as they are (instead od x(1), x(2), ... m x(19)) and if it's possible to get all solution vectors with their variable names.
You don't have to use an anonymous function handle as an ode function, you can create a separate function file (as shown in the odefun section of ode45.
For example, your odefun can look like:
function dy = myode(t,y)
% first unpack state variables
i_d2 = y(1);
i_q2 = y(2);
...
gamma2 = y(end-1);
omega2 = y(end);
% determine all constants
c34 = expression_for_c34;
...
c61 = expression_for_61;
% determine state derivative
i_d2_dot = expression;
...
omega2_dot = expression;
% pack state derivative based on order in state vector
dy(1) = i_d2_dot;
...
dy(end) = omega2_dot;
end
From this myode function you can also call other functions to e.g. determine the value for some coefficients based on the current state. Next, integrate the system using a suitable ode solver:
[t,y] = ode45(#myode,tspan,y0);
I'm solving a pair of non-linear equations for each voxel in a dataset of a ~billion voxels using fsolve() in MATLAB 2016b.
I have done all the 'easy' optimizations that I'm aware of. Memory localization is OK, I'm using parfor, the equations are in fairly numerically simple form. All discontinuities of the integral are fed to integral(). I'm using the Levenberg-Marquardt algorithm with good starting values and a suitable starting damping constant, it converges on average with 6 iterations.
I'm now at ~6ms per voxel, which is good, but not good enough. I'd need a order of magnitude reduction to make the technique viable. There's only a few things that I can think of improving before starting to hammer away at accuracy:
The splines in the equation are for quick sampling of complex equations. There are two for each equation, one is inside the 'complicated nonlinear equation'. They represent two equations, one which is has a large amount of terms but is smooth and has no discontinuities and one which approximates a histogram drawn from a spectrum. I'm using griddedInterpolant() as the editor suggested.
Is there a faster way to sample points from pre-calculated distributions?
parfor i=1:numel(I1)
sols = fsolve(#(x) equationPair(x, input1, input2, ...
6 static inputs, fsolve options)
output1(i) = sols(1); output2(i) = sols(2)
end
When calling fsolve, I'm using the 'parametrization' suggested by Mathworks to input the variables. I have a nagging feeling that defining a anonymous function for each voxel is taking a large slice of the time at this point. Is this true, is there a relatively large overhead for defining the anonymous function again and again? Do I have any way to vectorize the call to fsolve?
There are two input variables which keep changing, all of the other input variables stay static. I need to solve one equation pair for each input pair so I can't make it a huge system and solve it at once. Do I have any other options than fsolve for solving pairs of nonlinear equations?
If not, some of the static inputs are the fairly large. Is there a way to keep the inputs as persistent variables using MATLAB's persistent, would that improve performance? I only saw examples of how to load persistent variables, how could I make it so that they would be input only once and future function calls would be spared from the assumedly largish overhead of the large inputs?
EDIT:
The original equations in full form look like:
Where:
and:
Everything else is known, solving for x_1 and x_2. f_KN was approximated by a spline. S_low (E) and S_high(E) are splines, the histograms they are from look like:
So, there's a few things I thought of:
Lookup table
Because the integrals in your function do not depend on any of the parameters other than x, you could make a simple 2D-lookup table from them:
% assuming simple (square) range here, adjust as needed
[x1,x2] = meshgrid( linspace(0, xmax, N) );
LUT_high = zeros(size(x1));
LUT_low = zeros(size(x1));
for ii = 1:N
LUT_high(:,ii) = integral(#(E) Fhi(E, x1(1,ii), x2(:,ii)), ...
0, E_high, ...
'ArrayValued', true);
LUT_low(:,ii) = integral(#(E) Flo(E, x1(1,ii), x2(:,ii)), ...
0, E_low, ...
'ArrayValued', true);
end
where Fhi and Flo are helper functions to compute those integrals, vectorized with scalar x1 and vector x2 in this example. Set N as high as memory will allow.
Those lookup tables you then pass as parameters to equationPair() (which allows parfor to distribute the data). Then just use interp2 in equationPair():
F(1) = I_high - interp2(x1,x2,LUT_high, x(1), x(2));
F(2) = I_low - interp2(x1,x2,LUT_low , x(1), x(2));
So, instead of recomputing the whole integral every time, you evaluate it once for the expected range of x, and reuse the outcomes.
You can specify the interpolation method used, which is linear by default. Specify cubic if you're really concerned about accuracy.
Coarse/Fine
Should the lookup table method not be possible for some reason (memory limitations, in case the possible range of x is too big), here's another thing you could do: split up the whole procedure in 2 parts, which I'll call coarse and fine.
The intent of the coarse method is to improve your initial estimates really quickly, but perhaps not so accurately. The quickest way to approximate that integral by far is via the rectangle method:
do not approximate S with a spline, just use the original tabulated data (so S_high/low = [S_high/low#E0, S_high/low#E1, ..., S_high/low#E_high/low]
At the same values for E as used by the S data (E0, E1, ...), evaluate the exponential at x:
Elo = linspace(0, E_low, numel(S_low)).';
integrand_exp_low = exp(x(1)./Elo.^3 + x(2)*fKN(Elo));
Ehi = linspace(0, E_high, numel(S_high)).';
integrand_exp_high = exp(x(1)./Ehi.^3 + x(2)*fKN(Ehi));
then use the rectangle method:
F(1) = I_low - (S_low * Elo) * (Elo(2) - Elo(1));
F(2) = I_high - (S_high * Ehi) * (Ehi(2) - Ehi(1));
Running fsolve like this for all I_low and I_high will then have improved your initial estimates x0 probably to a point close to "actual" convergence.
Alternatively, instead of the rectangle method, you use trapz (trapezoidal method). A tad slower, but possibly a bit more accurate.
Note that if (Elo(2) - Elo(1)) == (Ehi(2) - Ehi(1)) (step sizes are equal), you can further reduce the number of computations. In that case, the first N_low elements of the two integrands are identical, so the values of the exponentials will only differ in the N_low + 1 : N_high elements. So then just compute integrand_exp_high, and set integrand_exp_low equal to the first N_low elements of integrand_exp_high.
The fine method then uses your original implementation (with the actual integral()s), but then starting at the updated initial estimates from the coarse step.
The whole objective here is to try and bring the total number of iterations needed down from about 6 to less than 2. Perhaps you'll even find that the trapz method already provides enough accuracy, rendering the whole fine step unnecessary.
Vectorization
The rectangle method in the coarse step outlined above is easy to vectorize:
% (uses R2016b implicit expansion rules)
Elo = linspace(0, E_low, numel(S_low));
integrand_exp_low = exp(x(:,1)./Elo.^3 + x(:,2).*fKN(Elo));
Ehi = linspace(0, E_high, numel(S_high));
integrand_exp_high = exp(x(:,1)./Ehi.^3 + x(:,2).*fKN(Ehi));
F = [I_high_vector - (S_high * integrand_exp_high) * (Ehi(2) - Ehi(1))
I_low_vector - (S_low * integrand_exp_low ) * (Elo(2) - Elo(1))];
trapz also works on matrices; it will integrate over each column in the matrix.
You'd call equationPair() then using x0 = [x01; x02; ...; x0N], and fsolve will then converge to [x1; x2; ...; xN], where N is the number of voxels, and each x0 is 1×2 ([x(1) x(2)]), so x0 is N×2.
parfor should be able to slice all of this fairly easily over all the workers in your pool.
Similarly, vectorization of the fine method should also be possible; just use the 'ArrayValued' option to integral() as shown above:
F = [I_high_vector - integral(#(E) S_high(E) .* exp(x(:,1)./E.^3 + x(:,2).*fKN(E)),...
0, E_high,...
'ArrayValued', true);
I_low_vector - integral(#(E) S_low(E) .* exp(x(:,1)./E.^3 + x(:,2).*fKN(E)),...
0, E_low,...
'ArrayValued', true);
];
Jacobian
Taking derivatives of your function is quite easy. Here is the derivative w.r.t. x_1, and here w.r.t. x_2. Your Jacobian will then have to be a 2×2 matrix
J = [dF(1)/dx(1) dF(1)/dx(2)
dF(2)/dx(1) dF(2)/dx(2)];
Don't forget the leading minus sign (F = I_hi/lo - g(x) → dF/dx = -dg/dx)
Using one or both of the methods outlined above, you can implement a function to compute the Jacobian matrix and pass this on to fsolve via the 'SpecifyObjectiveGradient' option (via optimoptions). The 'CheckGradients' option will come in handy there.
Because fsolve usually spends the vast majority of its time computing the Jacobian via finite differences, manually computing a value for it manually will normally speed the algorithm up tremendously.
It will be faster, because
fsolve doesn't have to do extra function evaluations to do the finite differences
the convergence rate will increase due to the improved precision of the Jacobian
Especially if you use the rectangle method or trapz like above, you can reuse many of the computations you've already done for the function values themselves, meaning, even more speed-up.
Rody's answer was the correct one. Supplying the Jacobian was the single largest factor. Especially with the vectorized version, there were 3 orders of magnitude of difference in speed with the Jacobian supplied and not.
I had trouble finding information about this subject online so I'll spell it out here for future reference: It is possible to vectorize independant parallel equations with fsolve() with great gains.
I also did some work with inlining fsolve(). After supplying the Jacobian and being smarter about the equations, the serial version of my code was mostly overhead at ~1*10^-3 s per voxel. At that point most of the time inside the function was spent passing around a options -struct and creating error-messages which are never sent + lots of unused stuff assumedly for the other optimization functions inside the optimisation function (levenberg-marquardt for me). I succesfully butchered the function fsolve and some of the functions it calls, dropping the time to ~1*10^-4s per voxel on my machine. So if you are stuck with a serial implementation e.g. because of having to rely on the previous results it's quite possible to inline fsolve() with good results.
The vectorized version provided the best results in my case, with ~5*10^-5 s per voxel.
In my script, I call the ODE solver ode15s which solves a system of 9 ODE's. A simplified structure of the code:
[t, x] = ode15s(#odefun,tini:tend,options)
...
function dx = odefun(t,x)
r1=... %rate equation 1, dependent on x(1) and x(3) for example
r2=... %rate equation 2
...
dx(1) = r1+r2-...
dx(2) = ...
...
dx(9) = ...
end
When reviewing the results I was curious why the profile of one state variable was increasing at a certain range. In order to investigate this, I used conditional debugging within the ode function so I could check all the rates and all the dx(i)/dt equations.
To my big surprise, I found out that the differential equation of the decreasing state variable was positive. So, I simulated multiple rounds with the F5-debug function, and noticed that indeed the state variable consistently decreased, while the dx(i)/dt would always remain positive.
Can anyone explain me how this is possible?
It is not advisable to pause the integration in the middle like that and examine the states and derivatives. ode15s does not simply step through the solution like a naive ODE solver. It makes a bunch of calls to the ODE function with semi-random states in order to compute higher-order derivatives. These states are not solutions to system but are used internally by ode15s to get a more accurate solution later.
If you want to get the derivative of your system at particular times, first compute the entire solution and then call your ODE function with slices of that solution at the times you are interested in.
I am trying to solve two equations with complex coefficients using ode45.
But iam getting an error message as "Inputs must be floats, namely single or
double."
X = sym(['[',sprintf('X(%d) ',1:2),']']);
Eqns=[-(X(1)*23788605396486326904946699391889*1i)/38685626227668133590597632 + (X(2)*23788605396486326904946699391889*1i)/38685626227668133590597632; (X(2)*23788605396486326904946699391889*1i)/38685626227668133590597632 + X(1)*(- 2500000 + (5223289665997855453060886952725538686654593059791*1i)/324518553658426726783156020576256)] ;
f=#(t,X)[Eqns];
[t,Xabc]=ode45(f,[0 300*10^-6],[0 1])
How can i fix this ? Can somebody can help me ?
Per the MathWorks Support Team, the "ODE solvers in MATLAB 5 (R12) and later releases properly handle complex valued systems." So the complex numbers are the not the issue.
The error "Inputs must be floats, namely single or double." stems from your definition of f using Symbolic Variables that are, unlike complex numbers, not floats. The easiest way to get around this is to not use the Symbolic Toolbox at all; just makes Eqns an anonymous function:
Eqns= #(t,X) [-(X(1)*23788605396486326904946699391889*1i)/38685626227668133590597632 + (X(2)*23788605396486326904946699391889*1i)/38685626227668133590597632; (X(2)*23788605396486326904946699391889*1i)/38685626227668133590597632 + X(1)*(- 2500000 + (5223289665997855453060886952725538686654593059791*1i)/324518553658426726783156020576256)] ;
[t,Xabc]=ode45(Eqns,[0 300*10^-6],[0 1]);
That being said, I'd like to point out that numerically time integrating this system over 300 microseconds (I assume without units given) will take a long time since your coefficient matrix has imaginary eigenvalues on the order of 10E+10. The extremely short wavelength of those oscillations will more than likely be resolved by Matlab's adaptive methods, and that will take a while to solve for a time span just a few orders greater than the wavelength.
I'd, therefore, suggest an analytical approach to this problem; unless it is a stepping stone another problem that is non-analytically solvable.
Systems of ordinary differential equations of the form
,
which is a linear, homogenous system with a constant coefficient matrix, has the general solution
,
where the m-subscripted exponential function is the matrix exponential.
Therefore, the analytical solution to the system can be calculated exactly assuming the matrix exponential can be calculated.
In Matlab, the matrix exponential is calculate via the expm function.
The following code computes the analytical solution and compares it to the numerical one for a short time span:
% Set-up
A = [-23788605396486326904946699391889i/38685626227668133590597632,23788605396486326904946699391889i/38685626227668133590597632;...
-2500000+5223289665997855453060886952725538686654593059791i/324518553658426726783156020576256,23788605396486326904946699391889i/38685626227668133590597632];
Eqns = #(t,X) A*X;
X0 = [0;1];
% Numerical
options = odeset('RelTol',1E-8,'AbsTol',1E-8);
[t,Xabc]=ode45(Eqns,[0 1E-9],X0,options);
% Analytical
Xana = cell2mat(arrayfun(#(tk) expm(A*tk)*X0,t,'UniformOutput',false)')';
k = 1;
% Plots
figure(1);
subplot(3,1,1)
plot(t,abs(Xana(:,k)),t,abs(Xabc(:,k)),'--');
title('Magnitude');
subplot(3,1,2)
plot(t,real(Xana(:,k)),t,real(Xabc(:,k)),'--');
title('Real');
ylabel('Values');
subplot(3,1,3)
plot(t,imag(Xana(:,k)),t,imag(Xabc(:,k)),'--');
title('Imaginary');
xlabel('Time');
The comparison plot is:
The output of ode45 matches the magnitude and real parts of the solution very well, but the imaginary portion is out-of-phase by exactly π.
However, since ode45's error estimator only looks at norms, the phase difference is not noticed which may lead to problems depending on the application.
It will be noted that while the matrix exponential solution is far more costly than ode45 for the same number of time vector elements, the analytical solution will produce the exact solution for any time vector of any density given to it. So for long time solutions, the matrix exponential can be viewed as an improvement in some sense.
I have two arrays:
E= [6656400;
13322500;
19980900;
26625600;
33292900;
39942400;
46648900;
53290000]
and
J=[0.0000000021;
0.0000000047;
0.0000000128;
0.0000000201;
0.0000000659;
0.0000000748;
0.0000001143;
0.0000001397]
I want to find the appropriate curve fitting for the above data by applying this equation:
J=A0.*(298).^2.*exp(-(W-((((1.6e-19)^3)/(4*pi*2.3*8.854e-12))^0.5).*E.^0.5)./((1.38e-23).*298))
I want to select the starting value of W from 1e-19
I have tried the curve fitting tools but it is not helping me to solve it!
Then, I selected some random values of A0=1.2e9 and W=2.243e-19, it gave me a better results. But I want to find the right values by using the code (not the curve fitting Apps)
Can you help me please?
A quick (and potentially easy) solution method would be to pose the curve fit as a minimization problem.
Define a correlation function that takes the fit parameters as an argument:
% x(1) == A0; x(2) == W
Jfunc = #(x) x(1).*(298).^2.*exp(-(x(2)-((((1.6e-19)^3)/(4*pi*2.3*8.854e-12))^0.5).*E.^0.5)./((1.38e-23).*298));
Then a objective function to minimize. Since you have data J we'll minimize the sum-of-squares of the difference between the data and the correlation:
Objective = #(x) sum((Jfunc(x) - J).^2);
And then attempt to minimize the objective using fminsearch:
x0 = [1.2E9;2.243E-19];
sol = fminsearch(Objective,x0);
I used the guesses you gave. For nonlinear solutions, a good first guess is often important for convergence.
If you have the Optimization Toolbox, you can also try lsqcurvefit or lsqnonlin (fminsearch is vanilla MATLAB).