How does Vertex AI AutoML encode categorical variables?

105 Views Asked by At

How does Vertex AI AutoML handle categorical variables? I'm particularly interested in high cardinality categorical variables.

Relevant documentation from Google:

https://cloud.google.com/vertex-ai/docs/datasets/data-types-tabular#categorical-transf

With this cryptic line:

**Convert the category name to a dictionary lookup index and generate an embedding for each index. **

But they don't discuss how the embedding is generated? One-hot encoding? Target encoding? Other?

I have searched and read other documentation, but I haven't found an answer. I do know that BQML used to only use one-hot encoding

0

There are 0 best solutions below