Please forgive me if I've missed it, but neither here nor on the project website I could find a license or a statement that you don't want to offer the code under a license. I'd greatly appreciate it if you added a license.

Hi,
I tried to solve a particular problem with phasepack, however, it seems most of the algorithms failed to give a reasonable solution (but my homemade algorithm does, except it takes too long), I used the parameters given in the example folder. A typical set of data can be downloaded from my dropbox as a .mat file. I really appreciate if you can tell me what the problem is. Thank you.

I am using runImageReconstructionDemo
Here is my opt settings:
opts = struct;
opts.algorithm = 'PhaseLift';
opts.initMethod = 'optimal';
opts.tol = 1e-3;
opts.verbose = 2;

Stacktrace:
Error using reshape
To RESHAPE the number of elements must not change.

Error in runImageReconstructionDemo/A (line 120)
im = reshape(pixels,[numrows,numcols]);

Error in solvePhaseLift (line 91)
A = A(eye(size(x0,1)));

Error in solvePhaseRetrieval>solveX (line 214)
[sol, outs] = solvePhaseLift(A, At, b0, x0, opts);

Error in solvePhaseRetrieval (line 163)
[sol, outs] = solveX(A, At, b0, x0, opts); % Solve the problem using the specified
algorithm

Error in runImageReconstructionDemo (line 74)
[x, outs, opts] = solvePhaseRetrieval(@A, @at, b, numel(x), opts);

Description: After initialization is finished, the function A is expecting a vector of size nm x 1, but instead it gets nm x nm.

I understand the the generalized phase-retrieval problem solves |Ax| = b where both A and x are complex. But can phase-pack be used when A is complex and x is real?

In

muFinal = opts.regularizationPara;   % Regularization Parameter

but later in the code we define

mu = norm(eig(AtL(b0)),1)*opts.tol;

And then we use mu everywhere,
so I don't see how muFinal is useful in this code.

Are you planning to include one of the most popular versions of the PR algorithm: Hybrid Input-Output/Error Reduction (see ref. J. R. Fienup, "Reconstruction of an object from the modulus of its Fourier transform", doi.org/10.1364/OL.3.000027)?

I am confused about the actual use of this code, here is an excerpt from the source code:

% Phase retrieval can only recover images up to a phase ambiguity.
% Let's apply a phase rotation to align the recovered image with the
% original so it looks nice when we display it.
rotation= sign(recovered_image(:)'image(:));
recovered_image = rotationrecovered_image;

where "image" is the complex amplitude to be recovered. However, in the actual experiment "image" is unknown, so how should we modify the code in this case? By the way, simply commenting out these two lines of code does not work.

I would be very grateful if anyone could answer my doubts. @tomgoldstein @rohanchandra30