Feature Engineering in Deep Learning

What are the “feature engineering” tasks that can be carried out in deep learning problems, specifically Computer vision, NLP and Collaborative tasks ?
Also wondering if “feature engineering” is not applicable to deep learning problems since hidden structures are already identified in the process ?

For unstructured data, feature engineering is rarely helpful (except, of course, for data augmentation). It depends on the domain however - for instance, ECG data still requires feature engineering at this stage.

Structured data generally benefits a lot from feature engineering.

1 Like