Figure 7.

A representative example of uniform manifold approximation and projection (UMAP) of autoencoder embeddings from the testing set. The top picture depicts gold-standard data projections, and the bottom picture depicts predictions of the model. The shown model used 700 annotated examples in each classification category for training and achieved an F1-score of 0.91 evaluated on the out-of-sample test set. The classes are ordered as follows: 0-Power line noise, 1-Artifacts, 2-Pathological Activity, 3-Physiological Activity.