Is Kaggle Learn a “Faster Data Science Education?”
Kaggle Learn is "Faster Data Science Education," featuring micro-courses covering an array of data skills for immediate application. Courses may be made with newcomers in mind, but the platform and its content is proving useful as a review for more seasoned practitioners as well.
Kaggle Learn bills itself as "Faster Data Science Education," a free repository of micro-courses covering an array of "[p]ractical data skills you can apply immediately."
As I'm sure you are well aware, there are all sorts of free and low-cost data science education alternatives available via numerous online platforms. So why am I feeling it necessary to write about another data science learning resource?
As I plan to embark on a fresh fall learning initiative — once Those Lazy-Hazy-Crazy Days of Summer are out of my system — I wanted to first find some concise review material for concepts I have previously learned and skills I have already acquired but which may have gone a bit rusty on me. To be clear, Kaggle Learn does not bill its micro-courses specifically as review material; however, I am so far finding that they fit this requirement for me rather well (though, admittedly, I'm still early in the process).
Below is list of the current micro-course offerings from Kaggle Learn. I'm (more or less) starting at the top and working through the list in order before setting off on new learning adventures.
Kaggle Learn isn't exactly brand spanking new; it launched in January 2018 and has seen growth of 1500% since (according to Kaggle Learn). Its short course approach to practical data skills, however, seems to set it apart from other resources, with courses running in the 3-8 hour range.
Curious to find out more about Kaggle Learn, I contacted the project's team lead, Dan Becker, and asked him a few questions. He obliged with the below insights into the platform, its content, and its future.
Matthew Mayo: What was the motivating factor behind the micro-courses, as opposed to more traditional, lengthier online offerings?
Dan Becker: We're unique in a couple ways.
First, we focus more on practical applications instead of academic theory. When I was in data science consulting, I worked on projects solving business problems for dozens of companies. It's striking how conventional courses focus on so many skills that don't matter in practice, while missing many of the things that matter most. For example, students might learn to code an algorithm from scratch (which no one will ever need you to do again), but you won't learn how to manipulate data to get it into mainstream libraries that implement the algorithm.
I've also been involved in hiring many times, and I know how well interesting, hands-on projects can get a hiring manager's attention. So, we make our courses really short, so you can get to making projects sooner. Personal projects also help you focus on the skills involved in real work. As an another benefit, most people enjoy their personal projects, so they find more time to do it... whereas conventional courses can get boring and drag on for months.
So each course on Kaggle is designed to be the fastest possible way to start doing a new type of data science project.
Strategically, Kaggle is in a unique position. Most business charge for their courses. The prices for online courses are extremely fair (a small fraction of what you'll earn when you get a data science job), and the people creating them are putting in a lot of work to make something. But when someone needs to charge for their courses, they feel pressure to create longer courses. It would be hard to charge hundreds or thousands of dollars for a 4 hour course and then tell the user "your better off working on a personal project at this point." Since our courses are free, we don't feel this pressure to pack them with anything extra.
Matthew Mayo: Where do you see Kaggle Learn going from here? Do you have other micro-courses on the horizon? Will there be alternatives to micro-courses at some point?
Dan Becker: For someone who prefers longer courses, there are already a lot of great options. I'll steer them to one of the other great online platforms recreate than trying to create another version of the same thing. Here's what I'd warn this person though: You are going to start learning a lot faster once you get a job in the field, since you'll be learning on the job rather than just in your spare time. And you the greatest asset to getting that job is a portfolio of personal projects. So delaying your work on personal projects may delay the fastest part of your learning.
We're about to release a feature engineering course, natural language processing course, geospatial analytics course, and a reinforcement learning course. Each course will only be about 4 hours long. You won't become an expert in any of these fields in that amount of time. But you'll learn enough to work independently, find new answers as you need them, and start doing interesting work.
I've started with the machine learning and Python courses as refresher — there is so much about Python I don't use regularly — and will hopefully check them off the list one by one by the end of the month. Off to new material at that point, after shoring up my base.
I encourage anyone looking for short practical data skills courses to have a look at Kaggle Learn, whether you are a beginner or more of an intermediate looking for a solid set of review materials. While these courses won't make you an expert in any area in a few hours — and, indeed, they don't purport to do so — they will provide a concise path to practical implementation and a way forward from there.
- What 70% of Data Science Learners Do Wrong
- Kaggle Kernels Guide for Beginners: A Step by Step Tutorial
- What’s wrong with the approach to Data Science?