Comparison of different autoencoder variations vs state-of-the-art methods on real data channel 17, the optimal value of neighbours for Isomap was found as 20 through grid search.

The metrics indicate that the AE variant is able to outperform the conventional feature extraction algorithm with the exception of the CHS metric, while the performance of the other variants varies across metrics due to the fact that the internal metrics evaluate the quality of the clustering from different perspectives as no ground truth is available for real data. The values indicating the highest performance for each metric have been bolded.