The elements of my 5-dimensional dataset belong to one class only. What would be an appropriate method to decide if a new element is in that class, and how to use that method?
Is your one class really one class? You ask “if a new element is in that class”, that makes it a 2 class problem already. In class vs not in class, like is cat vs is not cat and so on. If it really is 1 class (i understand this to mean all observations are the same class), then isn’t that essentially unlabeled data?
Autoencoders can be used to reconstruct the input, and if reconstruction error is too high, the input is seen as different from objects the model trained on, useful for measuring dissimilarity/outlier detection without labels
I have gained from Wikipedia that an autoencoder is a neural network that learns to copy its input to its output. A new element means that I can add another sample (5 distinct numbers) to my set. There are also 44 references on “autoencoders” mentioned in the Wiki which I do not have time to scan them all. As I have access to Mathematica, I wonder if somebody has experience with autoencoders.