Anyone have any recommendations for tabular data transformers, particularly those that create embeddings optimized for anomoly detection? Going to be receiving only about 6 fields per observation, some of which will be categorical. Hugging Faces available models would be the best, but I will take what I can get 😄

Some things that popped on my radar and I've not deeply experimented with: https://huggingface.co/microsoft/table-transformer-detection https://github.com/microsoft/TUTA_table_understanding https://github.com/microsoft/table-transformer

Hmmm those look mainly focused on detecting tables in a document not trying to understand the patterns of the data in the tables.

Ah, that is true!

Yeah - there are a lot of good articles on NN/transformers for tabular data but I can't seem to find many pre-trained models