Ending fminsearch on a whim - matlab

Id like to be able to cancel fminsearch (eg using ctrl+c) and retaining the best solution so far. Is there a way to do this? I dont know exactly what sort of tolerance I want to set, I'd much rather "play by ear" and ctrl+c when I'm satisfied with the solution. Matlab keeps going onnn and on and on for what seems to be no gain.

fminsearch can be terminated early by both output functions and plot functions. To get an interactive demand from the user, use of graphical interfaces that don't wait until code is finished executing to update is basically essential, so using a plot function would be the most straightforward way of achieving this. The default plot function figure has "Stop" and "Pause" buttons built in, as seen in the documentation's example for calling:
fminsearch(funfcn,x,optimset('PlotFcns',#optimplotfval))
If the Stop button is clicked, fminsearch returns the result of its latest iteration and by default displays the information:
Optimization terminated prematurely by user.

Related

Matlab GUI windowButtonMotionFcn Slow When No Action Taken

I have a MATLAB (r2016a) GUI with a windowButtonMotionFcn callback. Inside the callback is an if-elseif-else block that changes the cursor depending on which axes in the GUI the mouse is over. It runs very quickly when it encounters the cases when it needs to change something, but according to profiler, takes a significant amount of time on the "return" command.
Can anyone offer insight into a way to correct this? It seems to me if it's not executing any more code, it should be running more quickly when it is exiting the function (i.e. not getting hung up on a "return") than when it has more code to execute.
A basic outline of the code is as follows:
function mouseMove(handles)
xy = %Get cursor position
if %xy over axes1
set(gcf,'Pointer','crosshair')
elseif %xy over axes2
set(gcf,'Pointer','arrow')
else
return %Here is where MATLAB is spending a lot of time
end
%A lot of additional code for when the cursor is over axes1 or axes2.
end
This may be related to the fact that the end of a function often appears in the profiler as taking a lot of time. In general, the profiler is good for identifying inefficiencies in your code (functions that are called too many times, etc.) but really isn't well-suited to actually benchmarking code since JIT acceleration isn't enabled. For benchmarking, it's best to use timeit or some other function.
That being said, when I implement your same callback, I don't see any performance hit for the return statement
As #Ander pointed out, you should pay attention to the Calls column as it's possible that the number of times that return is called is simply more than the rest of the callback. The default coloration of a given line is based on the total time and not the time per call.

MATLAB command line output not following

There is something very simple that can be very upsetting occasionally.
I sometimes want to keep track of the outputs of an algorithm at each iteration, so within a for loop, I use disp command of MATLAB to output some information. However, although there is quite a bit of time between the calling of dispcommands, the MATLAB command window falls behind and I need to use the mouse to keep up with it manually. Do you know if there is any way to have it not fall behind?
Thanks,
It really shouldn't fall behind, but I could see this happening if the computations are intensive and the MATLAB JVM, which drives the GUI, does get the resources to update the command window.
The following command may be of help:
drawnow update % or just drawnow
According to the documenatation page, this "causes only user-interface objects to refresh, if needed". If that fails, try just drawnow with no arguments to see if it helps to flush the entire system event queue, including graphics updates.
If that doesn't work you could insert a pause(0.01) or something similar as a last resort.
It shouldn't fall behind if you don't use the mouse at all. However, I often use the waitbar for this purpose.
FEX also has several text & GUI progress bars.

How to get the iteration number of lsqnonlin

I am doing parameter estimation in matlab using lsqnonlin function.
In my work, I need to plot a graph to show the error in terms of lsqnonlin iteration. So, I need to know which iteration is running at each point of time in lsqnonlin. Could anybody help me how I can extract the iteration number while lsqnonlin is running?
Thanks,
You want to pass it an options parameter setting 'display' to either 'iter' or 'iter-detailed'
http://www.mathworks.com/help/optim/ug/lsqnonlin.html#f265106
Never used it myself, but looking at the help of lsqnonlin, it seems that there is an option to set a custom output function, which gets called during every iteration of the solver. Looking at the specification, it seems that the values optimValues.iteration and optimValues.fval get passed into the function, which is probably the things you are interested in.
You should thus define your own function with the right signature, and depending on your wishes, this function prints it on the command line, makes a plot, saves the intermediate results in a vector, etc. Finally, you need to pass this function as a function handle to the solver: lsqnonlin(..., 'OutputFcn', #your_outputfun).
The simple way to do this would be:
Start with a low number of (maximum) iterations
Get the result
Increase the number of iterations
Get the result
If the maximum iterations is used Go to step 3
This is what I would recommend in most cases when performance is not a big issue.
However, if you cannot afford to do it like this, try edit lsqnonlin and go digging untill you find the point where the number of iterations is found. Then change the function to make sure you store the results you need at that point. (don't forget to change it back afterwards).
The good news is that all relevant files seem to be editable, the bad news is that it is not so clear where you can find the current number of iterations. A quick search led me to fminbnd, but I did not manage to confirm that this is actually used by lsqnonlin.

diagnostic for MATLAB ODE

I am solving a stiff PDE in MATLAB using ode15, and it often freezes depending on the initial conditions. I never actually get an error, it just won't finish even after 10 hours when it should take around 30 seconds to run. I am experimenting with different spatial and time node intervals, but it is hard, because I don't get feedback.
Is there some sort of equivalent to diagnostic for fsolve? stats is not useful because it only displays an output after fsolve is finished.
Check out the documentation on odeset, and specifically the stats option. I think you basically just want to set stats to on and you will get some feedback.
Also, depending on your ODE, you may need a different solver. About half way down the page on this page there is a list of most of the solvers available in MATLAB. Depending on whether your function is stiff or non-stiff, and how accurate you need to get, one of those might work better for you. Sometimes I just code them all in and comment out all but one until I find the one that runs the best for me, but check out the documentation on each if you want to find the "right" one for your application.
Your question is confusing because you refer to both ode15s and fsolve locking up. These are two completely different functions. One does numerical integration and the other solves for roots. Also, fsolve has no option called 'Stats' (see doc fsolve). If you want continuous output from fsolve use:
options = optimist('Display','iter');
[x,fval,exitflag] = fsolve(myfun,x0,options)
This will display the iteration count, number of function evaluations, the function value, and other stuff depending on what algorithm you use (the alorithm can be adjusted via the 'Algorithm' option). Again see doc fsolve for full details.
As far as the 'Stats' option with ode15s goes, it's not going to give you very much information. I doubt that it will you figure out why your system is halting (if it even is ode15s that you have a problem with). What you can try is using an output function via the 'OutputFcn' option of odeset. You can try the simple odeprint first:
options = odeset('OutputFcn',#odeprint)
which will print your state after each integration step. Type edit odeprint to see the code and how you might write your own output function if you need to do more.

Disable plots in Matlab

I have some programs written in Matlab that I need to run several times for some reasons (debugging, testing with different input, etc...)
But, there are a lot's of graphs that are plotted by the programs and its various functions such that everytime I run the program, I have to wait for all the graphs to be displayed, which is very annoying and time consuming (especially when you are working with a small laptop).
After the program is executed, I close them with a close all.
So my question is:
Is there a way to disable all plots/figures/graphs in Matlab? either in the options, or by executing a certain code like disable plot and enable plot to ensure that no figures are being displayed.
I know that I can just browse the code and comment the plotting part, but I don't want to forget to uncomment.
Try some combination of the two commands:
set(gcf,'Visible','off') % turns current figure "off"
set(0,'DefaultFigureVisible','off'); % all subsequent figures "off"
The second one, if you put it near the beginning of your program, might do the trick for you. Of course, it is still creating the plots, which might be undesirable for computation time and/or RAM issues.
This is a classic reason to avoid Matlab when one can. It fosters bad programming design. To solve this problem correctly, you should create something that lets you "flip a switch" at the highest level of your program and control whether plots show or do not show. Perhaps it even has gradations of the show/don't show option so you can select different types of plots that do/do not show depending on what diagnostics you are running.
Ideally, you'd want this "flip a switch" creation to be a class that has access to visibility and plot functions of other objects. But because interactive object-orientation is so cumbersome in Matlab, it's often not worth the effort to develop such a solution, and most people don't think about this design aspect from the outset of their project.
Matlab would encourage someone to solve this by making flag variables like "isPlotVisible" or something, and creating functions that always accept such flags. I agree this is a bad design.
You could run matlab from the command line with:
matlab -nojvm
but then you don't get the GUI at all. Alternatively, you could write a file 'plot.m':
function h = plot(varargin)
h = [];
end
which doesn't do anything. If this is in the working directory (or somewhere else near the top of the path), then plot will call your function instead of the 'real' plot. You'd need to do the same from any other graphing functions you call.
The closest way I know of 'turning off plotting' would be a folder of such functions that you can add to the path to disable plotting, and remove to enable.
The previous methods are fine, but an easy and good habit to take is to use a "on/off parameter". So basically, at the beginning of your code, you can add something like:
DisplayFigure = 1; %1 = display, 0 = no display
After that, add "if DisplayFigure == 1 ... end" for all your plotting related commands, where the commands should be inside the if statement (the ... above). Hence you won't even compute the plots, which will save you a lot of time and memory. You just have to change the value of the variable "DisplayFigure" to plot or not the figures.