ODE45 randomly returns vector of length 1.

I am solving ODEs with a script I run many times, and often what is happening is that I can run it several times sucessfully while changing parameter values every time, but then all of the sudden I get the error "script returns a vector of length 1,..." and then I can't run the script again because I keep getting this error. This has happened me seemingly randomly multiple times, and I can usually shut my computer down and either the next day or two, the error just stops happening even though I do not make any changes to the code.

Is this something that happens to other people, and am I missing something obvious? I could include my script if that helps, but this error happens to me with almost every ODE script I have to run multiple times while changing parameter values.

Thanks for any help.

EDIT: My files are included in the comments. When I call the ODE file, I get the error that the returned vector is of length of 1, but when I just compute the dxdt explicitly from the ODE file, I obtain a vector of length 3 so I'm not sure where the problem is.

[Cesar García Echeverry]

Can you share your code? Maybe is a problem definition.

[Noah Bolohan]

Here is my code:

clear all

%HLO Mutual Invasion

global b B a d f m h v u g

B=0.0625; %(0.0049,0,09)

a=0.3; %(0.4,2.7)

d=0.08; %(0.004,0.6)

h=0.07; %(0.33,1) maybe equal to g

u=0.5; %(0.07,0.9)

g=0.07; %(0.33,1) maybe equal to h

v=1/3; %v>0 in scenario 1

b=0.3;

f=3.2;

m=1.7;

sres=0.74;

Jump=0.01; %Grid step size

RosMacCond=f-m-b*(f+m); %>0 for stable steady state in xy system, <0 for periodic orbit

Tinit=800;

options = odeset('RelTol',1e-8,'AbsTol',[1e-6 1e-6 1e-6]);

Tmax=500;

PC=0; %Percentage trackers

PCold=0;

[TT,SS]=meshgrid(Jump:Jump:1-Jump,sres*Jump:sres*Jump:sres);

[J,I]=size(TT);

MU=zeros(size(TT)); %Floquet multiplier

OIC=zeros(size(TT));

for i=1:I

for j=1:J

Ts=TT(1,i);

s=SS(j,1);

clear X T

x0=0.5;

y0=0;

z0=0.5;

[T,X]=ode45(@HLOode,[0 Tmax],[x0,y0,z0],options);

xbar=X(end,1);

zbar=X(end,3);

LIC(j,i)=f*xbar/(b+xbar)-m;

PCold=PC;

PC=floor(100*((i-1)*J+j)/(I*J));

if PC==PCold+1

disp(['Lynx InvCond (xbar=',num2str(xbar),', RMC=',num2str(RosMacCond),'): ',num2str(PC),'% done.'])

end

end

end

LIC=LIC;

figure(1)

contour(TT,SS,LIC,[0 0],'m')

hold on

contour(TT,SS,LIC,[0.3 0.3],'m--')

hold on

axis([0 1 0 sres])

PC=0; %Percentage trackers

PCold=0;

Tinit=3000; %If RosMacCond<0, run xy0 for Tinit to get on orbit, then extract orbit during Tfinal time

Tfinal=300;

if RosMacCond>0

xbar=m*b/(f-m); %xbar value in xy system

disp(['xbar=',num2str(xbar),'. RM Condition = ',num2str(RosMacCond)])

OIC=TT.*(h.*xbar^2./(B+xbar^2)+SS./v-u)+(1-TT).*(g.*xbar./(d+xbar)-u); %Owl Invasion Condition

figure(1)

xlabel('T')

ylabel('s')

clear title

titlestr=['Invasion Conditions (fp). B=',num2str(B),', b=',num2str(b),', d=',num2str(d),', g=',num2str(g),', h=',num2str(h),', m=',num2str(m),', u=',num2str(u),', v=',num2str(v)];

title(titlestr)

hold on

contour(TT,SS,OIC,[0 0],'g')

hold on

contour(TT,SS,OIC,[0.1 0.1],'g--')

hold on

axis([0 1 0 sres])

elseif RosMacCond<0

IndexPeaks=0;

for i=1:I

for j=1:J

Ts=TT(1,i);

s=SS(j,1);

x0=0.5;

y0=0.5;

z0=0;

clear X T

[T,X]=ode45(@HLOode,[0 Tinit],[x0,y0,z0],options);

x0 = X(end,1); %Get on periodic orbit and take endpoint

y0 = X(end,2);

z0 = X(end,3);

while length(IndexPeaks)<2 %Run system until we get >1 full orbit. Increase Tfinal until we do

Tfinal=Tfinal+100;

clear X T

[T,X]=ode45(@HLOode,[0 Tfinal],[x0,y0,z0],options);

Tinter=linspace(T(1),T(end),100*Tfinal);

Hinter=interp1(T,X(:,1),Tinter);

Linter=interp1(T,X(:,2),Tinter);

[HPeaks,IndexPeaks]=findpeaks(Hinter);

end

PerStart=Tinter(IndexPeaks(1));

PerEnd=Tinter(IndexPeaks(2));

Period=PerEnd-PerStart;

CycleT=Tinter(IndexPeaks(1):IndexPeaks(2));

CycleH=Hinter(IndexPeaks(1):IndexPeaks(2));

CycleL=Linter(IndexPeaks(1):IndexPeaks(2));

A=Ts.*(h.*CycleH.^2./(B+CycleH.^2)+s./v-u)+(1-Ts)*(g*CycleH./(d+CycleH)-u);

MU(j,i)=(1/Period)*trapz(A);

PCold=PC;

PC=floor(100*((i-1)*J+j)/(I*J));

if PC==PCold+1

disp(['Owl InvCond (Periodic Orbit, xbar=',num2str(xbar),', RMC=',num2str(RosMacCond),'): ',num2str(PC),'% done.'])

end

end

end

figure(1)

xlabel('T')

ylabel('s')

clear title

titlestr=['Invasion Conditions -(po). B=',num2str(B),', b=',num2str(b),', d=',num2str(d),', g=',num2str(g),', h=',num2str(h),', m=',num2str(m),', u=',num2str(u),', v=',num2str(v)];

title(titlestr)

hold on

contour(TT,SS,MU,[0 0],'g')

hold on

contour(TT,SS,MU,[10 10],'g--')

hold on

axis([0 1 0 sres])

else

disp('RosMacCond=0, zero eigenvalue in Hare-Lynx system.')

end

The function file HLOode is

function dxdt=HLOode(t,x)

global b B a d f m h v u g Ts s

dxdt=[Ts.*(x(1).*(1-x(1))-x(1).*x(2)./(b+x(1))-x(1).^2.*x(3)./(B+x(1).^2))+(1-Ts).*(-x(1).*x(2)./(b+x(1))-a.*x(1).*x(3)./(d+x(1))); f.*x(1).*x(2)./(b+x(1))-m.*x(2); Ts.*(h.*x(1).^2.*x(3)./(B+x(1).^2)+s.*x(3)./(v+x(3))-u.*x(3))+(1-Ts).*(g.*x(1).*x(3)./(d+x(1))-u.*x(3))];

end

Answer 1

[James Tursa]

Try clearing your variables before you run the script. It could be that you are inadvertently using a variable from a previous iteration.

[Noah Bolohan]

I put clearvars at the start of my script, the error persists.

Answer 2

[John D'Errico]

If it happens with almost every script that you run, though not immediately, then it is something you are doing that is strange, because that is not behavior anyone else ever sees. Passing variables around using globals is always dangerous as hell, because it make things difficult to debug. I'm not sure where the problem arises, but I would suggest using the debugger. Set it to trap out when an error occurs. Then look very carefully at the variables involved, at how the functions are called. You may need to wander around in the stack to figure out the problem, but this is certainly not anything I have ever seen or heard of. And since you do not appear to be doing anything overtly dangerous with memory, that means you have done something nasty in the code.

If I had to make a guess, I wonder if possibly globals are indeed causing your problem. clear all should kill them off. But you have created something strange here, and the most likely cause seems to be the heavy use of globals.

So you might add an additional call to clear at the beginning to test this out:

clear GLOBAL

Though I do see this claim in the help for clear.

clear ALL removes all variables, globals, functions and MEX links.

Using the debugger is the only viable way to track it down.

[Noah Bolohan]

Do you have a suggestion for how to call variables into a function file then, as I am doing with my ODE, without using global variables? I can't tell what the error is via debugging but I am not well versed with these MATLAB details yet...

I have an edit in my question which maybe specifies more what the error could be.

I took my ODE function file and put it at the end of my original script, and now the error is that 'Ts is an undefined function or variable', even though I define it in the if loop right above where I call ode45. Any idea what is going on here?