A Busy Dad’s guide to learn Machine Learning — Part 1 — the Journey
Story of Busy Dad of 2 kids and full time IT Architect job trying to acquire this skill — Challenges I faced and how I overcame them — It is possible
Another one — really
People have shared their stories of learning machine learning. However, I have not seen any story targeted to the audience of IT leaders who have spent more than 15 years of experience in the software engineering field and have a full-time job. For them, the biggest question is — Can I learn this with the time you have every day? I hope my story will resonate with those people or rather motivate them to become leaders in machine learning engineering.
About me
I have worked as an Enterprise eCommerce Architect for almost 15 years before becoming a Digital transformation leader since I found out I am also good at connecting business strategy with IT strategy. I got exposure to machine learning when I was helping my colleague to research a recommendation product for our company. After reading a few articles and understanding the potential of this field, I got hooked into this and I wanted to learn the fundamentals. Now after training myself for almost 9 months now and after completing two specializations in Coursera and one Nano-degree in Udacity, I think this is the right time to share my journey.
You cannot learn in 3 months- Set the right expectations
Oh Heck — No. If you are working a full-time job like me and have kids, the maximum dedicated time you can get is between 2 and ½ hours in a day and during the weekend maybe a total of 6 to 8 hours — so the total you can get is between 15 to 20 hours max in week with 15 most of time.
Expect that your starting weeks to be slow — I started to get my ‘Aha’ moments only after 3 months — you may get that before that depending on your effort but setting right expectations is the first thing you should do. Ph.D. folks are not only smart but they also spent countless hours in their particular field. So skills mastery depends on how much time you dedicating on that.
Another example of hurrying up things and falling flat — I bought “ The Hundred-Page Machine Learning Book” by Andriy Burkov on the 6th week of my study with the expectation that I will complete reading the book in one week and master machine learning. Man — I was so wrong at that time. However, after 5 months, it is such a lovely read.
So give yourself some time and stay away from undue expectations since that could easily frustrate you. I think if you can stay disciplined, you can be in a great position at the end of 6 months. Discipline — I know it is hard.
Doggedness is your primary skill, Intelligence is secondary — coming back again and again
You read it right — Intelligence is necessary since it helps us in speedy comprehension but it comes second to your determination. Often I hear this complaint from my friend or colleagues who also started a machine learning course that it is so boring — oh heck yes since our mind has become trained in thinking of solutions in iterative programming which is good in solving business problems. This is a different paradigm — here you are helping the machine to detect patterns based on the data provided to it.
There were days when I could not make any headway, I started to question my intelligence, my ability to learn. Believe me, you will go through a similar situation but if you wanted to learn this skill keep coming back again and again. In such situations, I noticed giving yourself a break for 2–3 days work wonders.
There is a great article on Learning in HBR.org — “Learning is a learned behavior. Here’s how to get better at it”. So keep on learning even during excruciating boredom.
Another important thing celebrate every victory even if it is small — that was my chicken soup for my soul — my strength when it felt that I am just a dumb person and I didn’t have anyone to fall back to discuss. Learning by myself is a hard job. Then I started to celebrate every small advancement — Like ‘James Altucher — everyday 1% improvement’. You will need that since either you are learning in the early hours of morning or night when your family members are sleeping.
Math rules the world but you should not get scared by it
It is true and you have to revisit some statistics, vector operations, and derivative calculus but you should not get bogged down by the enormity. Following is a very interesting excerpt from the book “ AIQ by Nick Polson and James Scott.
“ There’s obviously a lot of sophisticated math behind these suggestion engines. But if you are math-phobic, there’s also some very good news. It turns out that there’s really only one key concept you need to understand, and it’s this: to a learning machine, ‘personalization” means ‘“conditional probability.”
To learn that one thing behind the machine learning algorithms. At first, you will struggle but soon your brain will support you in that endeavor.
To motivate you — Can you imagine that the world most patterns can be solved by one giant approximation function? Read this chapter when you start learning deep learning —
Do not get lost
There are so many courses that are teaching machine learning and often choosing the right one is a big task. So try to start with a course that teaches fundamentals and stick to it — like Math for machine learning, classical machine learning algorithms with python.
Often we select a course thinking how soon this can get us a job and I made the same mistake. Instead, select the course by how it can help you to learn the subject. You would get these types of courses in Udacity and Coursera and also Edx. I am not going to recommend any specific course since we all have our style of learning but at least you should start with basic mathematics used in machine learning.
So your path should be — Maths -> Classical Machine Learning → Deep Learning ( Follow up story in part 2 ).
You will fall in love with Python
And fall in love with programming again — at least I did. I am assuming that as an IT leader or architect you have programmed in java or c# but python will make your adrenaline going. And why not — It is the most preferred language to teach kids programming. It supports object-oriented, scripted and functional programming. You can quickly test your code in Jupyter notebook and I can go on. I know it is not strongly typed but who needs that.
You will have to read about data and business of data — constant reading is key in learning any subject
If you have similar IT experience like me, you have designed tables to store data or API data structures for rest integration. But for machine learning, you will have to look at data differently since that is the raw material — their relations and representations.
For this, you have read books — One book I can recommend to start with — ‘Numsense! Data Science for the Layman — Annalyn Ng and Kenneth Soo’ but there are others too. You also have to read business books on Applied artificial intelligence — my strategy was finishing one book every 3 weeks but as you can understand I was not successful every time. You need secondary reading to complement your training courses — a lot.
Summary
I hope that I will be able to motivate you to learn about this interesting field — We are only touching the surface in this industry and there will be so much more. Just think once you have learned this skill, you would be able to help your team to discover a drug for incurable disease or help your company to position products for the right audience and so many potential world problems you can solve. So start small and aim for big.
Thanks for reading my story and all the best for your learning. Please let me know what you think about my story so that I can follow up with the second one.
Second Part of my story
