Skip to main content
. 2023 Jul 19;23(14):6507. doi: 10.3390/s23146507
Algorithm 1 CNN-based Feature Selection.
  • 1:

    procedure CNNFeatureSelection(trained_cnn, f)

  • 2:

        Create a new model to output the activations of the last convolutional layer

  • 3:

        Compute the activations for the testing data X

  • 4:

        Set the desired number of features n

  • 5:

        Find the indices of the features with the highest mean activations

  • 6:

        and save the output in top_feature_indices variable

  • 7:

        Initialize an empty list as top_feature_names =[]

  • 8:

        Initialize a counter variable counter = 0

  • 9:

        for each index in top_feature_indices do

  • 10:

            if index < length(feature_names) then

  • 11:

                if counter < n then

  • 12:

                    top_feature_names ← f[index]

  • 13:

                    counter←counter + 1

  • 14:

            else

  • 15:

                continue

  • 16:

        return top_feature_names