This trains the network N times (reseting parameters and initializing after each iteration) and computes SHAP values for all trained models. It then aggregates this into one dataframe with mean and std. The 'value' columns is identical to values_mean and used for plotting
This trains the network N times (reseting parameters and initializing after each iteration) and computes SHAP values for all trained models. It then aggregates this into one dataframe with mean and std. The 'value' columns is identical to values_mean and used for plotting