Tensorflow datasets

[Ramin_Firouzi]

Hi,

In the course Data Pipelines with TensorFlow Data Services, week one, versioning datasets, the instructor says that if you specify “as_supervised” as true in the load function, then the dataset will be preformatted into tuples of data and label. If you set it as false, the dataset will be available as a dictionary. I know what are tuple and dictionary, but what is the advantage and disadvantage of having data in these two formats in this case?

[ai_curious]

Supervised learning requires a label associated with each record. Think

value value value cat
value value value not_cat

Setting the flag tells the service one of these columns is not like the others so it can be treated as the label. Otherwise, all columns are treated as ‘regular’ data.