Should you learn Data Science fast or slow?

29th April 2019



In the field of data science education, there seem to be 2 types of learners.

One is the fast learners 

This group focuses on learning “just in time” and only when they need to solve a problem. Their main metric of success is being able to get some actionable results even if they do not truly understand how the model works. This involves only general knowledge about a model: the parameters, the type of input data and interpretations of the output. 

You can recognise fast learners by their desire to use machine learning for every problem (even if a visualisation will do) and when ML is needed they only want to use a neural net (when a simpler model will do).

Fair criticisms are that they cannot explain the why behind problems or propose different approaches.

Second, is the slow learners

These learners focus on understanding math before they apply any of the techniques. They often come from academic backgrounds where this is the norm, and appropriately this group reads many textbooks and papers. 

You can recognize slow learners by wanting to explain the theory of a model when a result will do. They also are able to propose advanced techniques that are not known to others and a new combination of old techniques that are not built into sci-kit learn.

Fair criticisms are that learning this way is time intensive and not much is captured. Trying to master Andrew Ng’s course without knowing why you’re learning the course may lead to much of the information to be immediately forgotten.

I propose a 3rd Group the Fast & Slow Learners.

This group is small now but I believe it will grow. They care first about solving a specific problem, coming up with a solution, implementing with little knowledge and at the end learning the mathematics deeply. 

This makes the F&S learner understand why he is learning math. It gives the theory of context. It leads to higher retention and motivation.

This group never forgets about the math because if he does not understand it he will never be able to optimize a model. And when he learns it, it is done at a slow pace with the focus on understanding the why behind a model. Never memorize but intuit the model.

While most debate the importance of theory vs practice, realize that it is never a trade-off but the sequence is important. So as you learn first learn fast and then slow.

Read also: 7 tips to help you construct the perfect data science CV

Tagged with:

Fernando Hidalgo

A Data Scientist that has worked with Discovery Communications and Prismoji. Top Data Science writer in Quora with posts that have received over 370K views. He is passionate about building machine learning products and democratizing data science education.

To give you the best possible experience, this site uses cookies. Continuing to use jamieai.com means you agree on our use of cookies.
Accept