Honestly, either is fine. It depends on your learning style and priorities. If you want to jump quickly to getting state of the art results with rich datasets like images and natural language text, then start with the deep learning course. If you want to get a deeper understanding of feature engineering, model interpretation, and (especially) techniques for tabular data, start with the machine learning course.

Also, the machine learning course was originally the initial introduction for a masters program, so it is a little less intensive and fast moving. But it also assumes a little more math background in parts, since everyone in the program already was pretty familiar with linear algebra, probability, and statistics.

In the end, the two courses go together pretty well, so I don't think it matters too much what order you do them in.

> (especially) techniques for tabular data, start with the machine learning course.

Do you think DL techniques are going to better than traditional techniques for tabular data in the future?

This is in reference to your article: http://www.fast.ai/2018/04/29/categorical-embeddings/ and lessons 3 and 4 from the DL course.

Yes probably. For instance, there's been interesting recent examples from Pinterest and Instacart (amongst others) who found that DL-based approaches for tabular data reduced the engineering and maintenance headaches due to less need for manual feature engineering.

Although decision tree approaches maybe will continue to be faster to train in many cases. At the moment, it really helps to be familiar with both approaches.

Thank you for the clarification.

Great answer!.