How to save a workspace with fittypes that use anonymous functions?

I am defining fittypes that refer to an anonymous functions. Afterwards I fit the model and saved the workspace. Here is a minimalistic example:

testfunc = @(params1, params2, x) params1*x + params2*x.^2;
testfittype = fittype(@(params1, params2, x) 2*testfunc(params1, params2, x));
x = [1;2;3];
y = [1;5;20];
[model, gof, output] = fit(x, y, testfittype);

Warning: Start point not provided, choosing random start point.

Now whenever I close matlab and load the saved workspace I get warnings that seem to stem from the anonymous function definition inside the fittype.

Warning: Could not find appropriate function on path loading function handle

My question is how can I get around those errors? I know that I could just avoid putting an anonyous function inside a fittype definition but since my real examples are quite long I would like to avoid that.

[VBBV]

As the warning suggests, you need to load the saved workspace to the correct path where the userdefined function is present. You can set the path in preferences to load saved workspace automatically

[Sophie]

Thanks for your comment. I am a bit confused since the function is part of the workspace itself as it is testfunc that cannot be found. Could you elaborate a bit more? The complete warning is:

Warning: Could not find appropriate function on path loading function handle

/private/var/folders/5n/b6g8h5896bz8n3vqk32b3wym0000gn/T/Editor_pigjx/LiveEditorEvaluationHelperE1941464770.m>@(params1,params2,x)params1*x+params2*x.^2

> In uiimport/runImportdata (line 459)

In uiimport/gatherFilePreviewData (line 438)

In uiimport (line 260)

[Matt J]

As you can see, the code below works when run here in the online environment. So, whatever problem you're having, it is something about your environment locally.

testfunc = @(params1, params2, x) params1*x + params2*x.^2;
testfittype = fittype(@(params1, params2, x) 2*testfunc(params1, params2, x));

save test testfunc testfittype
clear

load test

x = [1;2;3];
y = [1;5;20];
[model, gof, output] = fit(x, y, testfittype)

Warning: Start point not provided, choosing random start point.

model = 
     General model:
     model(x) = 2*testfunc(params1,params2,x)
     Coefficients (with 95% confidence bounds):
       params1 =      -1.829  (-15.07, 11.41)
       params2 =       1.697  (-3.307, 6.702)

gof = struct with fields:
           sse: 3.3684
       rsquare: 0.9832
           dfe: 1
    adjrsquare: 0.9664
          rmse: 1.8353

output = struct with fields:
             numobs: 3
           numparam: 2
          residuals: [3x1 double]
           Jacobian: [3x2 double]
           exitflag: 1
      firstorderopt: 3.6964e-08
         iterations: 1
          funcCount: 6
       cgiterations: 0
          algorithm: 'trust-region-reflective'
           stepsize: 2.2520
            message: 'Success. Fitting converged to a solution.'
       bestfeasible: []
    constrviolation: []

[Sophie]

If I execute the save and load code in my local matlab I also do not get the warning. However if I close matlab and open it again, I do get the warning when importing the data from the previously saved workspace. Would you mind confirming that it works on your side when restarting matlab and then importing the previously saved workspace?

[Matt J]

The same code I ran aboe works on my local computer as well.

Answer 1

[SACHIN KHANDELWAL]

Hi @Sophie

It seems you're experiencing some difficulties while attempting to save a workspace that includes "fittype" function.

To help with this, I'd kindly suggest using MATLAB's "save" function to preserve your workspace in a '.MAT' file. Once saved, you can easily retrieve your workspace at any time by employing the 'load' function in MATLAB.

I tried it on my end, and it is working as expected.

Thanks!