Multi-label classification using Multi Task / Multi Model

[ghost]

I'm trying to create a model which has N categorical labels (multi-label) and each label has M classes (multi-class), the normal Structured Data Classification works with multi-class and multi-label right? The multi-class works with a specific number or inferring the number of classes, but multi-label its not clear how it works.

Can I use Multi Task / Multi Model to achieve multi-label classification? I saw the nodes, blocks and the example, but I didn't get any hints besides getting the Sigmoid Cross Entropy With Logits and putting it on the search space, which I'm not sure where and if it is possible aside from the nodes / blocks from Auto Keras.

[haifeng-jin]

The multi label feature is not like your use case. It is for case like a dataset with M classes and an instance may belong to multiple classes out of M classes.

[ghost]

Thanks, I get the idea after finish the prototype that I was working on. I read more on TensorFlow / Keras doc, I switched to Transformers, it makes way more sense for the problem that I need to solve, which has lots of labels and a lot of them are categorical. The Auto ML needed to validate cause the client wants to run it in a specific Cloud that offers it.