Skip to main content
. 2025 May 15;15(10):1255. doi: 10.3390/diagnostics15101255
Algorithm 1 Pseudocode of the alternating training.
  •   1:

    Inputs: Training data data =(xi,yi)

  •   2:

    Outputs: the Classification Model M and the Generator G

  •   3:

    Initialize parameters of both models

  •   4:

    for each epoch in epochs do

  •   5:

          for each batch datai in data do

  •   6:

                Freeze the parameters of the generator

  •   7:

                Calculate the loss value using the cross-entropy loss(Equation (2))

  •   8:

                Update the parameters of the classification model

  •   9:

          end for

  • 10:

          for each batch datai in data do

  • 11:

                Freeze the parameters of the classification model

  • 12:

                Calculate the reward

  • 13:

                Update the parameters of the Proximal Policy Optimization-Clip approach (Equation (3))

  • 14:

          end for

  • 15:

    end for