Becoming A Data Science Unicorn – Part 2

I hope you have already read the first part of this Series and found your strong WHY, if you have not – you know what to do. Also, if you didn’t go through the course I suggested there, AI for everyone, do so now, please. It will make your learning path incredibly smother since you will not stumble around unknown/misunderstood terms. If you already took the course, great job. I advise you to share it with your friends and family -it will make your life much easier and reduce the frequency of the incoming questions within the next few months: what on earth are you learning, why are you so excited, frustrated…

This time we will talk about how to actually start your data science learning path. Remember to hold to your why very tight, there is a lot of work in front of us (especially you!). Also, do not forget to have fun on your voyage.

Without further ado, let’s dive in. This post has three main parts:

1. Tips and tricks– some of the lessons I have learned on my learning path, the hard way.

2. Learning Approach – how to structure your learning path.

3. Resources – finally, I will share some of the learning resources I found helpful and those I wish I discovered/existed earlier.

Tips and Tricks

Comparing data science learning path with Moscow metro map

Don’t panic

a quick peek at the picture above will make you panic and run far away from this bottomless field or make you use all your awake time for learning – don’t do either. With endlessly emerging new resources I sometimes (read most of the days) feel the same. However, I learned one thing just in time. I learned it at the time I was simultaneously trying to follow 5 courses, read 8 books and numerous blogs (true story). At the time I got the book Rest: Why You Get More Done When You Work Less as a present and, of course, couldn’t find the time to read it. From this point of view, I know how my approach to learning may have seemed to those around me but the hype over all the “must know/do to be a data scientist” can mess with your head. Fortunately, at the time I was beginning to think I couldn’t do this (yea right), one simple quotation woke me up from my must hypnosis: “You can have it all, just not all at the same time.” (Betty Friedan). An even better one I can offer for your learning path is:“You cannot learn everything (nor should you) but you can get where you are heading by consistently learning one thing at the time.”

Time is your most precious resource

use it wisely. Somewhat connected with the first quote, one thing you should pay most attention to on your learning path is time. The best thing about the self-learning path is that you can choose what to learn and how to learn it. Use this advantage to maximize gained knowledge/skills and minimize invested time following three simple principles:

1. You have no time for ego – don’t waste time listening to some advanced course if you do not even understand the terms, it is perfectly ok to start with ‘for dummies’ resources.
2. You have no time for modesty and self-doubt – similar to the above, do not waste time on courses/resources that do not teach you anything new – it’s ok to admit you know something and move on.
3. You have no time for ‘must’ – do not waste time on resources just because you read somewhere that it is a must. If it does not suit your learning style or simply does not suit you, move on. There are so many resources out there – do not be afraid to find what suits you the best, take full advantage of making your own learning plan – your own.

Take the Breadth First Approach

do not go too deep into every topic as you will end up going down a rabbit hole. For your first curriculum, it is enough to cover the basics. In time, you will go more in depth for certain topics. Also, try to avoid the biggest time trap (at least for me) – spending too much time on theory and different theory sources, it is perfectly ok for theory to follow practice sometimes.

Take care of yourself

dear data science wizards, you are entering a life of learning on a daily basis. You cannot learn well in the long run if you are not in good physical and mental shape. So, try to take care of yourself, rest when needed, sleep enough. Do not face your learning as a speed race, prepare your body for a lifelong marathon.

Learning Approach

Let me tell you a secret. Even though most online courses give the opposite impression – there is so much more to a data science project than machine learning and there is so much more than technical skills to develop and improve on your path of becoming a data scientist. Skills like critical thinking, planning, time management, decision making, storytelling are the ones which will determine whether your real world data science project will succeed.

The good news is that you – yes you! already have all of these skills. Even better news is that you will get to practice and improve them – for the rest of your life.

The easiest way to improve these skills is through real data science projects, which you should definitely do – we will get back to that soon.

Something I realized along the way was that I can maximize my learning endeavours if I plan my learning in the structure of a data science project.

So, maybe you can try to use it the other way around – plan your learning as a data science project so by learning you will also learn to plan a data science project ?.

I hope it will became a bit clearer once I lay down the structure:

Problem & Goal definition

define the aim of your learning, trying to be as specific as possible. If your aim is a (the) data science position, be aware that data scientist has different definition and responsibilities in different companies. Make sure to learn about the different roles.  If you do not have a specific role in mind I suggest you choose the full stack data scientist or generalist for your study path. I believe in the following mantra in data science:
&nbsp
‘Becoming the jack of all trades will make you a better master of the one of your choosing.’

Identify available resources

put down your initial condition (your current knowledge), determine how much time you have to meet your goal and how much money you can invest (do not worry, you can create great learning path with free resources).

Domain knowledge

the domain for this specific problem is YOU. You should discover everything regarding your learning style: are you a top down or bottom up learner, do you learn best alone or in a group, with reading, listening or visual material… You will discover some of these things along the way but try to analyze as much as possible in advance, including all the possible setbacks, situations that can make you want to quit your learning and develop contingency plans. As in data science projects – all of this will lead you faster to your goal.

Divide and conquer

break the problem down into subproblems, determine even more specific topics to learn and skills to gain. It can be an iterative process where you define larger skill groups and then subgroups in each of those. It would be good if you could complete the smallest tasks within two weeks.

For each skill/topic/task:

1. Define an evaluation metric and threshold – determine how you will test your knowledge and skills;
2. Research – find possible resources and decide on the top few;
3. Analyze – narrow the previous list down and decide on your final learning plan;
4. Learn – acquire knowledge/skill using the style most convenient for your domain knowledge;
5. Evaluate – test your knowledge/skill;
6. Adjust approach – if you fail at 5, analyze what has gone wrong in your learning and improve it;
7. Repeat 4 and 6 until you reach the evaluation threshold.

Hope the learning structure will work for you and prepare you for your future data science projects. However, you should definitely modify it in the way you find useful and exciting. If you’re excited about your learning approach and having fun, you’ll make fast progress.

Learning Resources

Finally, here are some resources and the learning path I would personally use if I was about to start with data science. Do not be frightened by the number of them, I will list the ones I found most useful, feel free to cherry pick only a few or one per topic, based on the available resources you determine.

Coding

Start by learning/improving Python or R. I am personally team Python, so my advice might be a bit biased but here it goes: If you already know one of them, stick to it. Otherwise, choose Python.Coding is best learned by doing but here are some resources that might help.

• The Hitchhiker’s Guide to Python
• Think Python book 
• Real Python Tutorials 
• Pandas Python library cookbook
• Numpy Python library tutorial

Data Analysis & Data Wrangling

These skills are best learned by doing so get your hands dirty with data, you can find some using a Dataset Search. Some other cool practicing resources include:

• DataCamp and
• Dataquest

Linear algebra & Calculus

If you need to brush up your knowledge, here are some of my favorite resources:

• 229 Linear Algebra Notes
• The Matrix Calculus You Need For Deep Learning

If you need a bit in-depth knowledge, or even if you do not but have time, I highly recommend amazing courses by Gilbert Strang:

• Linear Algebra
• Matrix Methods in Data Analysis, Signal Processing, and Machine Learning

Probability and statistics

• Statistical Thinking for the 21st Century
• Introduction to Probability and Statistics
• Practical Statistics for Data Scientists: 50 Essential Concepts

Machine Learning

• The hundred-page machine learning book
• Machine learning course by one and only Andrew Ng

Deep Learning

• Hacker’s guide to Neural Networks
• Grokking Deep Learning
• Deep Learning book
• Deep Learning Specialization
• Fast.ai courses

If you are interested in specific Deep Learning domain:

• Convolutional Neural Networks for Visual Recognition
• Natural Language Processing with Deep Learning
• Deep Unsupervised Learning

Linux

• Linux mastery course

Data Visualization and Storytelling

• Storytelling with data book and resources
• Matplotlib Cookbook

Projects – practice, practice, practice

Although this one is one of the last listed, it is definitely not the least important, you should introduce it to your learning path as soon as possible. Try to work on projects you are personally interested and invested in. Here are some project inspirations:

• Kaggle
• DrivenData
• TunedIT
• Projects for different levels

Awesome resources

You can find lots of awesome resources searching awesome + ‘topic of your interest’ on GitHub. For example:

• Awesome data science
• Awesome machine learning

Wrapping up

There you go. We covered some of the tips & tricks, introduced a learning path structure and listed some resources. I hope these thoughts and resources will be helpful on your learning path.

However, if there is only one thing you will take away from this post it should be this one:

Do not follow anyone else’s learning path (my own included), especially not blindly.

You should create your own path, based on your goals and resources, using other people’s data (suggestions/advices) with scientific methodologies. Ultimately, voyaging on this one-of-a-kind learning road, with skills learned and knowledge acquired using your specific learning style, doing projects with personal motivations, tackling specific obstacles, conquering specific goals with all off your amazing uniqueness, you will soon start shaping into a one-of-a-kind Data Scientist – ?anyone?

Happy learning!