Table 3.

Evaluated hyperparameters for the experimental setup

Hyperparameter	Evaluated Values
Number of Principal Components	{32, 64, 128}
MLP Hidden Layers	{2, 3}
MLP Hidden Layer Size	{16, 32, 64}
Autoencoder Hidden Layers	{2, 3}
Autoencoder Hidden Layer Size	{16, 32, 64}
Masked autoencoder (37)	{Yes, No}
OC-SVM Kernel	{Linear, Radial Basis Function}
Number of Nearest Neighbors (N)	{10, 20, 50}
Final Refinement Neighbors	{10, 20}
kNN Graph Neighbors	{20, 50, 100}

The best hyperparameters, shown in bold, were chosen using cross validation on the known classes (whose labels are available) as described in (36).