How To “Ultralearn” Data Science: deep understanding and experimentation, Part 4
In this fourth and final part of the ultralearning data science series, it's time to take the final steps toward developing a deep understanding of the fundamentals and learning how to experiment -- the two aspects that are the ultimate keys to ultralearning.
By Benedict Neo, Data Science enthusiast and blogger.
- How To “Ultralearn” Data Science, Part 1, What is "ultralearning" and how can you follow the strategy to become an expert of data science?
- How To “Ultralearn” Data Science: removing distractions and finding focus, Part 2
This is the last part of the series on ultralearning data science. In this last segment, I will be discussing deep understanding and experimentation.
With the skills to optimize learning to the maximum, it’s time to enact strategies to cultivate profound comprehension and be open to experimentation that will engender innovation.
In each case, I will be using Richard Feynman and Vincent van Gogh respectively as examples.
Cultivating Deep Understanding
Physicist Richard Feynman was known for his uncanny intuition; he had the knack of looking at a complex problem and seemingly plucking the solution out of thin air.
The technical term for this ability is intuitive expertise, and it can seem rather mysterious to outside observers.
But there’s a perfectly rational explanation for Feynman’s flashes of brilliance: his deep understanding of physics enabled him to intuit unexpected connections and patterns.
“I learned very early the difference between knowing the name of something and knowing something.”
― Richard P. Feynman
To be a perspicacious data scientist, it takes time and patience to build up the level of deep understanding on which intuitive expertise is built.
But by employing these four strategies, you can accelerate the rate at which you acquire it.
(1) Get the basics right
Start by getting back to basics. Feynman was famous for asking “stupid questions” and would frustrate his students by bombarding them with questions about basic concepts.
However, Feynman knew something his students had yet to learn: it’s impossible to progress to complex concepts when you only have a vague understanding of foundational concepts.
On the other hand, it’s impossible to become an intuitive expert until you know the foundational concepts of your field inside-out.
Data science requires heavy mathematics and machine learning.
However, that does not mean you have to memorize each theory and fact by heart. In reality, there are tons of libraries and packages that allow you to apply maths and ML to your project.
All you need is a basic understanding of the concepts and insight into what they are. So, make sure you master the foundational concepts, and you’re all set!
(2) Take the longer, harder route
A challenging learning experience can lead to a deeper grasp of the subject.
That’s why you should try and embrace the struggle. Resist taking shortcuts in your learning; if there are two ways to arrive at a solution, choose the longer, more involved one.
Instead of using clean and well-structured data for your projects, get a little dirty, and obtain unstructured data from the internet and start cleaning it.
This will prepare you to deal with any kind of data thrown at you in the future. In your next data science project, try to use raw data, and clean it yourself. You’ll learn a lot.
(3) Persevere on difficult obstacles
Try not to give up immediately when things get challenging. Instead, implement a struggle timer. Force yourself to sit with every challenge or obstacle for at least 10 minutes before you look for a simpler solution.
Asking the right questions can be immensely difficult as you require creativity, analysis, a lot of research, etc. Basically, solving a problem requires you to sit down and start brainstorming with a pen and paper, albeit a little old fashioned.
In lieu of jumping to conclusions and rushing into the solution, contemplate them and weigh in all the factors involved.
(4) Replicate concepts for yourself
Finally, deepen your understanding of core concepts by replicating them for yourself. Look at the ideas and processes that expert practitioners in your field have formulated, then try and prove them or replicate them for yourself.
In other words, you are understanding the procedure and thought patterns behind it. This proffers you an opportunity to pick their brains and helps you to achieve deep knowledge and intuitive expertise.
To pick the brains of a dexterous data scientist, you could read their articles, work, or blogs that give you a peek into what their workflow and process are like.
For example, you watch a YouTube video on how to use TensorFlow to perform object detection. After watching it attentively, you replicate the entire process and apply TensorFlow to your model with your own personalized functionality and use case.
How did Vincent van Gogh go from an art school dropout, whom classmates recalled as an “unremarkable” painter, to the innovative artist who painted masterpieces like Sunflowers and Starry Night?
Through sustained, relentless experimentation. Look back over van Gogh’s full oeuvre, and you’ll see he didn’t hit on his distinctive aesthetic immediately.
Instead, he tirelessly tried different styles and techniques until he mastered his craft. Then, he experimented even more, finally arriving at a unique style.
Experimentation is ultralearning’s secret ingredient — the technique can take you from an accomplished practitioner to a true innovator. But, experimentation can seem a little overwhelming at first.
Importance of experimentation in data science
“Experiments are designed to identify causal relationships between variables, and this is a really important concept in many fields and particularly relevant for data scientists today.”
In The key to agile data science: experimentation, it says:
“The nature of data science is experimental. You don’t know the answer to the question asked of you — or even if an answer exists. You don’t know how long it will take to produce a result or how much data you need. The easiest approach is to just come up with an idea and work on it until you have something.”
In the blog titled To be a more effective data scientist, think in experiments by Aleksey Bilogur:
“The fundamental unit of value in data science is the experiment.”
The art of data science is the art of generating good hypotheses.
A “good” data scientist knows which avenues they can explore that are likeliest to actually have an impact on model performance. They are able to make hypotheses and run experiments that have the highest probability of actually improving their model performance.
This “sense” of what the “right” thing to do is comes from experience first, from understanding the dataset second, and from technical expertise third.
Being a good “artist” translates into being a good worker: you spend less time trying out dead ends, and more time making valuable improvements to your models.
With these, it is evident that the underlying notion of data science is about experimentation.
A scientist uses the scientific method to collect empirical evidence in many experiments relating to a hypothesis, to support or contradict a theory.
A data scientist, fundamentally, is a scientist who experiments with data continuously until the model (theory) is accurate(valid).
Three steps for experimentation
- Copy then create
If you’re wondering where to begin, one technique you can use is to copy then create: emulate someone else’s work, then use this as a stepping-stone towards testing your ideas.
“Good artists borrow, great artists steal.” — Pablo Picasso
Another thing you can do to jump-start your experimentation is to impose some constraints on it.
This might seem counter-intuitive, but limiting your creativity can help it blossom. That’s because working within strict limits can help you shake off your working habits and force you to try something new.
Finally, aim for the unexpected by hybridizing your materials, techniques, or skills to find your hidden superpower. Combining two seemingly disparate elements can lead to great results.
For example, if you have a background in astrophysics and you have a passion for art and music, you can use machine learning to visualize a black hole, create simulations, and even predict what planets and stars from billions of light-years away look like.
Using packages like astroML and deep learning, the sky’s the limit. With vigorous experimentation and creativity, and hybridizing your skillset, you can create groundbreaking work and achieve the impossible.
To truly become a data scientist, one has to develop a profound understanding of data and be bold to experiment with different models. Both these qualities will ensure the success of extracting hidden insights from data and engender valuable information.
Four steps you should take today:
- Master the basics.
- Take the longer route and persevere on difficulty.
- Borrow from experts then create.
- Experiment with different tools and skillsets with hybridization.
Original. Reposted with permission.
- How To “Ultralearn” Data Science, Part 1
- How To “Ultralearn” Data Science: removing distractions and finding focus, Part 2
- How To “Ultralearn” Data Science: optimization learning, Part 3