Balancing Machine Learning Theory with Practice
Theory and practice are complimentary aspects of a career in data, particularly practitioners of machine learning. People often fall predominantly to one side of the theory-practice fence depending on their natural interests and learning style. But to be good in this field requires you actively make a point to strike a balance between the concepts we read and the real-world scenarios we face.
So how do you strike this balance?
Build what you Read!
Many people shy away from building things because it’s not comfortable. Constantly tracing down bugs, endlessly traversing stack overflow, trying to bring so many interacting parts together into one system. Reading on the other hand is fun and easy because the material flows like a perfect little story. All the concepts have settled into place and the narrative we are consuming is how we all wish things worked.
But if you’re not uncomfortable you’re not learning. This is of course true for both theory and practice, but I think it’s easier to get too comfortable with reading. We feel good when we can rip through journals and understand the concepts, but you won’t make a difference in your field by only reading about the accomplishments of others. If you build what you read, and fail (and you will fail) you will gain an understanding that cannot be absorbed through reading.
Remember, you are not building to make something work; you are building in order to run into the wall of frustration again-and-again. That wall is your teacher and the one true source of your career potential. Once you learn to volley theory and building back-and-forth you’ll see just how rewarding a balanced approach can be.