gradient descent method with Gerchberg–Saxton algorithm ?

Hello everyone, may anyone help me in using gradient descent method with Gerchberg–Saxton algorithm ?

the code for GS Algorithm is :

clc
close all
clear all
model=zeros(150);
model(5:20, 20:121)=1;
model(24:140, 40:50)=1;
model(24:140, 85:97)=1;
figure(1);imagesc(model); colormap(gray)
DS=model; 
model=2*pi.*rand(150)-pi; 
pi_signal1=double(imread('pi.bmp'));
pi_signal=rgb2gray(pi_signal1);
pi_signal=imresize(pi_signal, [150 150]);
j=fftshift(fft2(pi_signal));
j=abs(j)/(max(max(abs(j))));
j=imadjust(abs(j), [0; 0.01], [0; 1]);
for k=1:1:150; 
  step1=fftshift(fft2(model));
  phase=angle(step1); 
  Gu=medfilt2(j).*exp(i*phase);
  gx=fft2(fftshift(Gu));
  model=pi_signal.*exp(i*angle(gx));
  figure(7),imagesc(abs(gx)); colormap(gray), title(num2str(k));
  err(k)=-sqrt(sum(abs(gx(1:k,1))-pi_signal1(1:k,1))^2/k);
  t(k) = 10*k;  %simulation time, where dt is time for one increment of loop 
end
figure(3)
plot(err,t);

Thanks

[Matt J]

You haven't mentioned what difficulty you are having.

[Zulqarnain Sheikh]

Thanks for response, I just want to implement gradient descent in GS algorithm, but I am unable to write the code of 2D gradient descent and I am unable to find one online. Any help will be much appreciated