Machine learning vs Reasoning and knowledge base: When to use what?

Hello guys,

in a lot of papers scientists use a knowledge base to check or represent data.

I am comparing machine learning versus reasoning systems, which use a knowledge base as a backeend. In my opinion machine learning wins over logic, when:

• Performance is an issue
• What is not in the knowledge base is not known to the reasoning system
• Unknown anomalies are completely invisible to the creator of the knowledge base

What do you think is an argument against/for using logic and a knowlege base versus machine learning?

I appreciate your reply!