using noise nmr sequences wavelet coefficient as a cnn-resnet input wish the corresponding nmr sequences as the cnn-resnet output to finish denoise in matlab

main.m
% Load the NMR sequence data and its corresponding noise data
load('NMR_data.mat')
load('Noise_data.mat')

% Apply the wavelet transform
wname = 'db4';
[C, L] = wavedec(NMR_data, 5, wname);

% Extract the wavelet coefficients of the NMR sequences
nCoeff = numel(C);
nSamples = L(1)+1;
X = zeros(nCoeff, nSamples);
for k = 1:nSamples
    indices = (k-1)*nCoeff+1:k*nCoeff;
    X(:, k) = C(indices);
end

% Normalize the wavelet coefficients and the NMR sequences
X = normalize(X);
NMR_data = normalize(NMR_data);

% Split the dataset into training, validation, and testing sets
train_ratio = 0.7;
val_ratio = 0.15;
test_ratio = 0.15;
[trainInd,valInd,testInd] = divideblock(length(NMR_data),train_ratio,val_ratio,test_ratio);

X_train = X(:, trainInd);
Y_train = NMR_data(:, trainInd);
X_val = X(:, valInd);
Y_val = NMR_data(:, valInd);
X_test = X(:, testInd);
Y_test = NMR_data(:, testInd);

% Define the CNN-ResNet model
input_shape = [nCoeff, 1];
output_shape = [nSamples, 1];
num_filters = 64;
filter_size = 3;
residual_blocks = 10;
l2_reg = 1e-4;

inputs = keras.layers.Input(shape=input_shape)

#first convolutional layer
x = keras.layers.Conv1D(num_filters, filter_size, padding='same', kernel_regularizer=keras.regularizers.l2(l2_reg))(inputs)
x = keras.layers.BatchNormalization()(x)
x = keras.layers.Activation('relu')(x)

#residual blocks
for _ in range(residual_blocks):
    y = keras.layers.Conv1D(num_filters, filter_size, padding='same', kernel_regularizer=keras.regularizers.l2(l2_reg))(x)
    y = keras.layers.BatchNormalization()(y)
    y = keras.layers.Activation('relu')(y)
    y = keras.layers.Conv1D(num_filters, filter_size, padding='same', kernel_regularizer=keras.regularizers.l2(l2_reg))(y)
    y = keras.layers.BatchNormalization()(y)
    x = keras.layers.add([x, y])
    x = keras.layers.Activation('relu')(x)

#last convolutional layer
x = keras.layers.Conv1D(num_filters, filter_size, padding='same', kernel_regularizer=keras.regularizers.l2(l2_reg))(x)
x = keras.layers.BatchNormalization()(x)
x = keras.layers.Activation('relu')(x)
x = keras.layers.Conv1D(1, filter_size, padding='same', kernel_regularizer=keras.regularizers.l2(l2_reg))(x)
x = keras.layers.BatchNormalization()(x)
predictions = keras.layers.add([inputs, x])

model = keras.models.Model(inputs=inputs, outputs=predictions)

# Compile the model
model.compile(loss='mean_squared_error', optimizer='adam', metrics=['mean_squared_error'])

# Train the model
epochs = 100
batch_size = 32
history = model.fit(X_train, Y_train, batch_size=batch_size, epochs=epochs, validation_data=(X_val, Y_val))

# Evaluate the model
score = model.evaluate(X_test, Y_test, verbose=0)
print('Test loss:', score[0])
print('Test accuracy:', score[1])

# Use the model to denoise the noisy NMR sequences
C_noise = wavedec(Noise_data, 5, wname);
X_noise = zeros(nCoeff, nSamples);
for k = 1:nSamples
    indices = (k-1)*nCoeff+1:k*nCoeff;
    X_noise(:, k) = C_noise(indices);
end
X_noise = normalize(X_noise);

Y_pred = model.predict(X_noise);

% Inverse the wavelet transform and obtain the denoised NMR sequences
denoised_NMR_data = waverec2(Y_pred, L, wname);
3136 chars
96 lines