While I absolutely agree with the core of this, deep learning can be a bit ‘arbitrary’ at times — for certain types of problems it will out perform all of the machine learning solutions. It really depends on what you’re trying to accomplish. For many of my aspirational tasks, there is no equivalent to a CNN, RNN, or DRN.

However if you need to crunch through arbitrary structured data, a random forest (or similar) is likely going to be your best friend.