How to Define Your Learning Path in Data Science
Deciding what to learn next is hard.
Data science is a booming field with an ever-increasing number of specializations and roles to fill. Along with the boom, there are more courses, books, and articles that are aimed towards helping data scientists learn the skills they need to be good at their jobs (or to get a job). There are so many things to learn. So what is the best way to decide where you should go next?
Starting with the basics might seem like a dumb thing to write about. I have been at the beginning of the journey into data science though. I know that deep learning seems super cool and sexy. That is not the place to start though. Make sure that you have a solid understanding of basic math, statistics, and computer programming before you try to learn anything crazy.
Once you have a solid understanding of the basic concepts that make up this thing we call data science, there are a lot of different options for how to decide what to learn next. There is no perfect way, and there are pros and cons to each learning path you might take.
Option 1: Focus Your Learning on What’s New
One way to structure your learning is to focus on keeping up with the state of the art in your chosen area. You can do this by reading and learning how to implement new research papers that come out. For many people this is one of the harder ways to learn because it requires you to read and understand research on a level deep enough to be able to translate it to code.
The benefit of focusing on the state of the art research is that you are always aware of the best known solutions to a lot of hard problems. It puts you in a position to work on interesting projects and build things that few, if any, others are building. It also sets you up to do your own research if that is interesting to you.
First, it is hard to read research papers. It takes time to comb through a research paper and understand everything that the researchers did. It is also difficult to understand exactly how useful the research is to the type of problems you are working on. I have found that many people do not enjoy learning this way, and if you don’t enjoy learning, you are not going to take the time to do it.
Option 2: Specialize in an Area of Interest
If there is a specific area of data science that interests you, such as natural language processing, computer vision, explainable machine learning, or compute-efficient deep learning, then you can just dive into that area. Becoming a specialist in something that interests you is a very valid way of approaching learning and it makes it fairly easy to know which things are worth learning to you.
The nice thing about learning this way is you are only studying what you are really interested in. If you focus your study in one specific area, it is also easier to become an expert on that subject matter, making you a valuable asset.
With the plus of becoming a specialist, comes the negative of becoming a specialist. Focusing all of your learning in one area can pigeonhole you into a position with less flexibility. You might miss out on some other valuable skills that a more general approach could offer you.
Option 3: Learn Things to Complete a Job or Project
Focusing on learning things that are going to be most useful to whatever you are currently working on is another great way to decide what to learn. If you are working on a project where time series modeling would be valuable you can start studying that. Once you finish that project, you find what topic will most help your next project.
This approach to learning helps you have a purpose to what you are studying. It gives you clear direction and motivation to your learning time. On top of that, there is a clear, immediate payoff to most of the things that you learn.
I have found that bouncing around between topics tends to leave some gaps in knowledge. For example, learning how to do image classification on the fly might cause you to miss the understanding behind how convolutions work. Sometimes these holes end up being trivial, but sometimes they are important holes you will have to go back and fill in later.
Option 4: Follow a Curriculum
This can be anything from a college degree, to an MOOC, to a book. Any sort of structured program that lays out what to learn and provides the materials to learn it.
One huge benefit of this is the convenience of it. You are told what is important to know and provided everything you need to become proficient at those skills. Most of the time you are provided with examples and exercises so that you have hands-on experience too.
Unfortunately, not all courses are created equal. If you choose to go this route you need to consider if the place you are getting your courses has a strong reputation and will provide good content. Another con of this path is that it is generally the most expensive of the options. You may also find that the skills and techniques you learn are not always easily applicable to the real world like they are in classrooms.
Option 5: Add Value to What You Provide
This way of deciding what to learn may actually fit under any of the other ones. It is the idea that from a career perspective, you are learning the things that will make you the most valuable. For example, if your company (or a company you hope to work for) values full-stack data scientists, a statistician might focus learning on dev ops.
Learning with the intent of making yourself more valuable will most likely lead to the biggest jumps in a career. It can also open a lot of different doors for new things to do.
This sometimes leads you to studying topics you are not interested in. This can make studying and mastering new material very difficult, and doesn’t inspire continued learning.
After you cover the basics, the rest of the options are not mutually exclusive. In fact, I think the best way to approach learning is some combination of all of these things. When it comes down to it, the best way to learn is finding a way of learning that you enjoy that you will stick to.
This post was originally posted on Towards Data Science.