give a code which uses back propagation network for character recognition in matlab

main.m
% Load the training data
load('character_data.mat');

% Set the parameters
inputLayerSize = size(trainingData, 2);
hiddenLayerSize = 25;
outputLayerSize = 26;
learningRate = 0.01;
numIterations = 1000;

% Initialize the weights and biases
weights1 = randn(inputLayerSize, hiddenLayerSize);
biases1 = randn(hiddenLayerSize, 1);
weights2 = randn(hiddenLayerSize, outputLayerSize);
biases2 = randn(outputLayerSize, 1);

% Training loop
for iteration = 1:numIterations
    % Forward propagation
    hiddenLayerInput = trainingData * weights1 + biases1';
    hiddenLayerOutput = sigmoid(hiddenLayerInput);
    outputLayerInput = hiddenLayerOutput * weights2 + biases2'; 
    outputLayerOutput = sigmoid(outputLayerInput);
    
    % Backpropagation
    outputLayerError = outputLayerOutput - trainingLabels;
    hiddenLayerError = (outputLayerError * weights2') .* (hiddenLayerOutput .* (1 - hiddenLayerOutput));
    
    % Gradient descent
    weights2 = weights2 - learningRate * (hiddenLayerOutput' * outputLayerError);
    biases2 = biases2 - learningRate * sum(outputLayerError, 2);
    weights1 = weights1 - learningRate * (trainingData' * hiddenLayerError);
    biases1 = biases1 - learningRate * sum(hiddenLayerError, 2);
end

% Testing
testData = ... % your test data here
hiddenLayerInput = testData * weights1 + biases1';
hiddenLayerOutput = sigmoid(hiddenLayerInput);
outputLayerInput = hiddenLayerOutput * weights2 + biases2'; 
outputLayerOutput = sigmoid(outputLayerInput);

% Obtain the predicted labels
[~, predictedLabels] = max(outputLayerOutput, [], 2);

% Display the predicted labels
disp(predictedLabels);
1623 chars
48 lines

main.m
function y = sigmoid(x)
    y = 1 ./ (1 + exp(-x));
end
56 chars
4 lines