
The one thing that all data scientists can agree on is that data science is hard to learn.
But what sets apart the learning experience of those who learned data science in 6 months from those whose journeys have dragged on for years?
The answer may come down to the method of learning they used.
While those who joined FAANG companies after only a few months of study will say that it was their dedication and tireless work ethic that helped them learn data science concepts, there may be more to the story than them putting in 80-hour learning weeks.
In fact, many of these successful data scientists likely used the Ultralearning method to efficiently learn everything they needed to in a short period of time.
What is Ultralearning?
Ultralearning is a learning method developed by Scott Young as a way to acquire skills and knowledge in a shorter period of time through self-directed, intensive study.
How is Ultralearning different from every other learning method out there? There’s proof that it works.
For example, Young used Ultralearning to learn the entire MIT undergraduate computer science curriculum in one year – a process that usually takes four years to complete. Not only that, but he passed all of the associated exams.
If that wasn’t enough, he did it again, though this time, his learning project was to not speak English for an entire year. Young and his friend Vat Jaiswal traveled to four countries over the course of one year to learn four languages. While they weren’t able to perfectly complete a year without speaking English, they both achieved an intermediate proficiency level in Spanish, Portuguese, Chinese, and Korean.
Ultralearning is an uncommon strategy due to its usage of aggressive learning tactics and highly self-directed nature.
However, when it comes to data science, using uncommon or unconventional learning tactics can be advantageous.
Who should Ultralearn?
It’s important to note that Ultralearning is not for everyone.
Ultralearning is not for those who are unwilling to push themselves to learn the difficult things first. Nor is it for those who aren’t comfortable with self-directed learning. It also isn’t for those who are more interested in "dabbling" – the process of picking up a skill now and then with no goal in mind of how to get better. In short, if your idea of learning data science is to passively watch tutorial videos without getting your hands dirty and writing code, then Ultralearning isn’t for you.
On the flip side, Ultralearning may be for you if you’re interested in learning something concrete through results-driven methods. Ultralearning is a great learning method if you’re willing to eliminate distractions, focus on your weaknesses, and get creative when what you’re doing isn’t yielding the results you need. This learning method is the essence of immersion – a way to fast-track yourself towards skill application and eventually mastery. In short, if you aren’t afraid of doing things the hard way, Ultralearning is for you.
How does Ultralearning benefit the data science learning experience?
As I mentioned in a recent article about microlearning, the data science learning process must be expedited.
The requirements for a data science professional are so great that the learning process must be as efficient as possible – otherwise, it will take years for anyone to transition into this burgeoning field. With additional competition coming from candidates with backgrounds in higher education, bootcamps, and online certificates, the learning process must also guarantee measurable growth towards skill mastery and application.
Enter Ultralearning, a method that guarantees success – if followed correctly. With Ultralearning, you are responsible for your success or failure. By designing a proper Ultralearning curriculum, you will end up jumping to an intermediate level of proficiency by quickly pushing through the difficult parts of being a beginner.
The 9 core principles of Ultralearning.
Young outlines nine steps for developing an Ultralearning strategy:
1. Meta-Learning
Meta-learning is what lays the foundation for your learning journey. In essence, it is the process of determining the most effective and efficient way to learn your desired subject or skill.
How this applies to learning data science: When it comes to meta-learning for data science, you need to do three things:
- Break data science down into its fundamental components.
- Figure out what your end goal is.
- Determine what has worked well for others learning data science in the past.
Data science, at its most basic level, is a discipline combining mathematics and programming. The entire data analysis process, Artificial Intelligence, and machine learning are all topics that expand on basic mathematics and programming.
When it comes to determining your end goal, you want to choose something specific. Whether you want to become a data analyst, develop machine learning skills, or just complete your first analysis, your end goal will help you build your learning curriculum.
Once you’ve broken Data Science down into its fundamental components and determined your end goal, you need to do your research into what worked well for others who embarked on the same learning journey. Towards Data Science and Youtube are two great resources for finding testimonials and advice on how to pursue your learning goal.
2. Focus
Focus is all about building up your ability to concentrate and eliminating distractions.
This is accomplished by scheduling dedicated distraction-free Ultralearning time into your schedule and using study techniques such as the Pomodoro method. To limit distractions, try working in a dedicated space, leaving your phone in a different room, turning off notifications, and using website blockers.
How this applies to learning data science: Start by scheduling one-hour blocks of dedicated Ultralearning time into your schedule. Then, divide that hour into a work period and a break period (for example, 50 minutes of work followed by a 10-minute break). The idea behind this is that by breaking the learning session into smaller chunks, information can be absorbed and comprehended more efficiently. Young doesn’t suggest breaking work periods into less than 30-minute blocks as it doesn’t promote the focus required for Ultralearning.
3. Directness
Directness is all about diving in head-first and learning the things you’re trying to become good at. This means getting your hands dirty and getting the hands-on experience right away.
How this applies to learning data science: When it comes to learning data science, this means getting your hands dirty by writing code, completing linear regressions and bivariate analyses by hand, and beginning your very own data science project.
Being direct can be intimidating and overwhelming, especially when it feels like there are too many things to learn at once. However, by breaking down the application into individual parts, it can be more feasible to attack each part one at a time.
For example, to complete a chi-square test for the first time, you would break it down into the individual steps of stating the null and alternative hypotheses, determining the critical value of chi-squared, computing the expected/observed value of chi-squared, and then concluding whether or not the null hypothesis should be accepted or rejected. Once you break down a complex topic into its individual parts, it becomes much more simple.
4. Drill
"Drill" is the concept of focusing on improving your weakest skills. Instead of avoiding practicing what you’re not good at, you need to consistently practice until you become good.
How this applies to learning data science: Start by creating a running list of your weakest skills. From there, break those skills down further into their individual parts. Then, begin practicing those individual parts to the point where you can put them all together to complete the skill as a whole.
A great example of this is the practice of mathematics. There are hundreds of practice problems online that you can complete over and over again until you become comfortable with a specific mathematical concept to the point where you can apply it to your data analyses or Machine Learning models.
5. Retrieval
Retrieval is all about testing yourself to figure out if you’ve learned something and to assess your skill level.
How this applies to learning data science: A great way to test your data science knowledge is to compete in Kaggle competitions or to complete coding challenges on Hacker Rank. By forcing yourself to apply what you’ve learned, you can identify any knowledge gaps or weaknesses in your foundation. The key is to not be afraid to make mistakes and to learn from any shortcomings.
6. Feedback
Feedback is what will make you a more well-rounded individual during and after the learning process is complete. By seeking out feedback on your learning, you can identify whether you’re improving and what areas need extra help.
How this applies to learning data science: Google, Stack Overflow, Kaggle, Reddit, Twitter, and Medium are all great platforms where you can receive feedback on your work. Stack Overflow is the perfect place to receive help and feedback on your code, and Reddit, Kaggle, and Medium are great places to receive feedback on data analyses or conceptual understanding.
7. Retention
Retention is all about making sure that you’re learning things to remember them. It’s important to be aware of what you’re forgetting and to figure out a way to make it stick.
How this applies to learning data science: Data science is a discipline that is best learned through building strong foundations. Those foundations can’t be built on top of each other if the one below it is structurally unsound. Therefore, ensure that you have a strong grounding in each topic before moving on to the next.
8. Intuition
Intuition involves developing a deep understanding of topics.
How this applies to learning data science: A great way to test whether or not you have a deep understanding of a topic is to try to teach it to someone else. This can be accomplished by using the Feynman Technique. Richard Feynman developed the technique when he discovered that upon creating a lecture for an undergraduate physics class that he couldn’t explain the topic at an undergraduate level. In short, if you can’t explain a concept in the simplest terms, you don’t understand it yourself.
9. Experimentation
Experimentation is all about expanding your horizons by exploring what you’ve learned outside of your comfort zone. This is done by trying to solve problems from different angles and exploring new ways of doing things.
How this applies to learning data science: Data science can always be done through brute force, which is usually what happens during the learning experience. However, once you’ve become competent in various data science areas, it’s now time to begin to refine and finesse your skills. Look for new ways to solve problems more efficiently, and try refactoring your code to make it more effective. Now is a great time to break open old projects and see how you can make them better with your newfound expertise. This is also a great time to begin contributing to open source projects or trying to apply your skills to real-life problems.
How to develop an Ultralearning project.
Young outlines three steps for developing an Ultralearning project:
- Pick what you want to learn.
- Choose the project format.
- Prepare to start learning.
1. Pick what you want to learn.
The key is to pick one thing that you want to learn deeply, intensely, and quickly.
Ultralearning projects must be specific, otherwise, they won’t work. Therefore, pick a specific skill. When it comes to data science, this could be data analytics, mathematics, Programming, machine learning, artificial intelligence, or more. However, because data science is such a broad field, it would be better to center an Ultralearning project around one specific thing, such as coding or mathematics.
Young highlights that shorter projects need more constraints. In other words, if your project is going to last only one month, then you need to ensure that constraints are in place to make sure that progress is noticeable. For example, if you are going to take one month to learn statistics, it’s better to focus on learning everything up to linear regressions instead of trying to take on the entirety of the statistics field.
Finally, Young also suggests that first-time Ultralearners shouldn’t set hard deadlines or goals for their project. This is because you’ll know quickly whether or not your goal is feasible which can lead to discouragement if you realize that your project is unfeasible in the timeline that you chose. A better course of action is to begin your learning project and then determine what your deadline or goal should be.
2. Choose the project format.
There are three different formats for an Ultralearning project:
- Full-time projects: Full-time projects are exactly that – full-time. Costly, intense, and fast, these projects are great if you can dedicate 40-hours per week to learning.
- Fixed-schedule projects: Fixed-schedule projects have concrete hours that you will devote to them each week. These hours are scheduled and occur at the same times every week. For example, you could dedicate 1 hour of learning every day before work or dedicate 2 hours of learning every day after dinner.
- Fixed-hour projects. Fixed-hour projects have a set number of hours that you complete whenever you find the time in your schedule.
Once the format is picked, you need to select a length of time. Young suggests that if your weekly time investment is low that your project should have a longer time period (such as 6–12 months) or a reduced scope (a more specific focus for the project).
3. Prepare to start learning.
The trick to a successful Ultralearning project is to take some time before learning to prepare.
This involves:
- Researching the skill you’re trying to learn
- Determining what has worked well for others in the past
- Gathering materials
- Researching different learning strategies
- Scheduling your time
- Conducting a pilot week of the schedule.
Young suggests spending no less than 50% of the length of the project on preparation.
Final thoughts.
Ultralearning is the perfect complement to the rigors of data science – intense, self-directed, and results-based.
For those who aren’t afraid of diving in head-first and doing things the hard way, Ultralearning can harness self-accountability and intensive study to yield incredible results in a short period of time.
For those looking to break into data science as a career change, this could be just the way to do it without spending copious amounts of time without guaranteed results.
By exploring alternative learning methods, the journey towards data science becomes open to all sorts of people, including those with no given background in the industry.
Ultralearning is extreme, to say the least, but with extreme practices can come amazing results. Why not see what Ultralearning can do for your data science journey?