Sklearn多任务：输入数据不是3维的吗？

Question

I have one huge data matrix X, of which subsets of rows correspond to different tasks that are related but also have different idiosyncratic properties. Thus I want to train a Multi-Task model with some regularization and chose sklearn's linear_model MultiTaskElasticNet function. I am confused with the inputs of fitting the model. It says that both the X and the Y matrix are 2-dimensional. The 2nd dimension in Y corresponds to the number of tasks. That makes sense, but in my understanding the X matrix should be 3-dimensional right? In that way I have selected which subsets of my data correspond to different tasks as I know that in advance (obviously).

Does someone know how to enter my data correctly for this scikit-learn module? Thank you!

Answer 1

I don't see why you'd want X to vary for each task: the point of multitask learning is that the same feature space is used to represent instances for multiple tasks which can be mutually informative. I get that you may not have ground truth y for all instances for all tasks, though this is currently assumed in the scikit-learn implementation.