How to learn Machine Learning 5 days a week, 9 months in a row
I studied machine learning every day for nine months and then got a job.
I quit Apple. Launched a web startup, but nothing came of it. My soul didn’t lie to this.
I wanted to learn machine learning. This is what inspired me. I was going to study everything to the smallest detail. I did not need to program all these rules, they will do everything for me. But I didn’t have a job.
Inspiration doesn't pay my bills.
On weekends, I started working part-time at Uber to pay for my studies.
I loved communicating with new people, but I hated being constantly behind the wheel. Traffic jams, brake, gas, constant thoughts about gasoline and refueling, air, air conditioning, gear shifting, where you can and cannot go, all these thoughts.
I studied machine learning. Day and night, five days a week. It was difficult. Yes, and now it’s not easier.
Uber on the weekend. Machine learning on weekdays. That was my everyday life. I had to study. I was obliged to learn, I just could not drive. At that time I did not have a life goal, but I knew for sure that this was not driving. One Saturday night I earned $ 280 and received a fine of $ 290. Minus $ 10 for one night.
Nine months after receiving my personal AI degree, I found a job. And it was the best job of my life.
How did I manage to practice every day?
1. Reduce your search space
Machine learning is a vast area. There are code, and mathematics, probability, statistics, data and algorithms.
There is no shortage of training resources. And the abundance of options equals their absence.
If you are serious about learning, then create a curriculum for yourself. Instead of spending weeks thinking about whether you should learn Python or R, start a course on Coursera or edX, start with math or code, spend one week developing a rough plan, and then stick to it.
This is how I came to my own master's degree in AI . I decided that I would first study the code and chose Python as the programming language. I searched everywhere for different courses and books and selected the ones that interested me the most. Is this method suitable for everyone? Most probably not. But he was mine, so it worked.
As soon as I had a curriculum, a path along which I could move, I could no longer waste time thinking. I could get out of bed, sit down at the table and learn what I needed (and what I wanted) to learn.
I was not strict with myself. If I found something interesting, I turned off my path and learned new things.
You should create your own path if you study online and not at university.
2. Change your environment
Your grandfather's first orange farm failed.
The soil was good. He planted the seeds. And the equipment could not fail.
It was too cold. Oranges grow at high temperatures. Your grandfather knew how to grow oranges, but the cold climate left them no chance.
He moved to a warmer city and again tried to open another farm.
Twelve months later, your grandfather’s orange juice became the best in the city.
Learning is like growing oranges.
Neither a laptop, nor the Internet, nor the best books will help you without the motivation to learn.
The problem is what surrounds you.
There are tons of things in your room that are so easy to get distracted.
You are trying to engage with friends, but they are not as dedicated as you.
Whatsapp messages arrive every seven minutes.
What to do about it?
I turned my room into a shelter for classes. Washed it. Turned off all alerts and put the phone in a chest of drawers in another room.
I warned my friend that I’ll talk to him after 16:00 when I turn on the phone. He agreed.
I love spending time with friends, but the study time is just for study.Can't hold out all day without a phone? Start with one hour. Any box that you don’t see will do. Do Not Disturb should be the default.
Change your environment and knowledge will flow towards you.
3. Set up the system so that you always remain the winner
Problem 13 put me at a standstill. I'm stuck.
I wanted to deal with her yesterday, but could not.
Now I need to study, but I worked so hard yesterday and nothing came of it.
I put off. I know that I have to study. But not today.
This is a loop.
Hell. I was already in a similar situation. However, nothing has changed.
A stack of books stared at me. Problem number 13. I start a timer. 25 minutes I know that maybe I won’t solve the problem, but I can sit down and try.
Four minutes have passed, I feel like in hell. It's just awful, but I'm not stopping. Twenty-four minutes later I don’t want to stop.
The timer goes off, I'm looking for a new one. And one more. After three approaches, I solve the problem. I tell myself that I am the best engineer in the world. Of course, this is not true, but it does not matter.
Even a small achievement is an achievement.
You cannot always control your academic progress. But you can control the time given to some things.
You can control: four sets of 25 minutes a day.
Unable to control: complete each new task on the same day.
Set up your system to always win.
4. Sometimes do nothing
I came to the following conclusion. Training - is the most important skill . If I can learn to learn better, I can do anything better. I can learn machine learning, I can become a good programmer, I can write better. I decided that I should improve my teaching methods. And immediately began.
I took a course on Coursera "Learning to learn" . One of the topics was focused and distracted thinking.
Focused thinking happens when you complete a single task.
Distracted thinking happens when you think of nothing.
At the intersection of these two ways of thinking, there is a moment for better learning. Therefore, it is in the soul that the best ideas and thoughts come to us. Because nothing else is happening there.
Broken thinking allows your brain to bind together all that it absorbed during focused thinking.
The catch is that in order for it to work correctly, you need to use both ways of thinking.
If you are doing four approaches of focused thinking for 25 minutes, then go for a walk. Take a nap. Sit down and think about what you learned.
As soon as you start to do nothing more often, you will see the value of many things because of the free space. A room is four walls around space, there is nothing but air in the bus, and the ship is sailing due to empty space.
There should be more to do in your training routine.
5. Sucks to put up with
Today you will learn what you forget tomorrow.
Then again, and forget again.
You spend all weekends studying, go to work on Monday and everything repeats.
Someone asked me how I remember for a long time what I read in books. I replied that in no way. With any luck, I will remember 1% of the total content. However, when this 1% overlaps with 1% of something else, magic happens. At such times, I feel like an expert on point connections.
You understand how much remains to be learned when you are already learning something throughout the year.
When will it end?
Never. You are always at the very beginning of the journey.
6. Three Years Principle
The other day I was in a park.
There was a little boy - he ran around the park, having a great time. Climbed up a hill and rolled down from it, ran for a tree and ran out because of it, into the mud, up and down the slope.
He laughed, jumped, and laughed again.
His mom came up.
“Come on, Charlie, we need to go.”
She led him away, and he continued to laugh, waving his blue plastic spatula.
What fascinated him so much?
He played. He had fun. The whole world was new. In our culture, there is a clear line between work and play. Learning is work.
You must study to get a job. You have to work to make money. For money, you can buy free time. And only in this free time can you be like Charlie - rushing and laughing.
If you perceive study as work, it will seem that you are in hell. Because you can always learn more. And you know the rule - only work, no games.
Now suppose that learning is a process from one topic to the next.
It’s like you are connecting different objects in a game.
And then you, like Charlie, going down the hill, will have the same feeling.
You learn one thing, use it to learn something else, get stuck, overcome it and learn something else. Turn the whole process into a dance.
I found out that if you have structured data in the form of tables, columns or data frames, then ensemble algorithms like CatBoost, XGBoost and LightGBM are best suited. And for unstructured data like pictures, videos, natural language or audio, your choice should be deep learning and/or transferring training.
I connected the dots. I told myself that the expert in this. Danced from point to point.
Exercise in this way, and by the end of the lesson you will have more energy than at the beginning.
This is a three-year principle. Take everything as a game.
I have to go to bed.
This is a bonus.
If you sleep poorly, you will study poorly.
You probably don't sleep enough.
I definitely did not sleep. Friday and Saturday evenings were the most profitable in Uber. People went to restaurants, parties, nightclubs. I - no, I drove. I worked until 2-3 in the morning, returned home and slept until dawn at 7-8 in the morning. Two days were a nightmare. It was Monday, and I was supposedly living in a different time zone. On Tuesday it was a little better, by Wednesday everything fell into place. But Friday came and everything repeated.
Such disturbed sleep patterns were simply unacceptable. I wanted to improve my training. Sleep cleanses the brain, allowing new connections to form. I finished work at 10-11 pm, returned home and slept for 7-9 hours. Less money, more knowledge.
Do not change sleep for learning. Do the opposite.
Machine learning is a vast area.
And in order to study well not only her, but also anything, you must remember:
- Reduce your search space
- Change the environment
- Resign yourself to screw it up
- Sometimes do nothing
- Think of learning as a game, and
- Sleep is the path to knowledge
Learn the details of how to get a sought-after profession from scratch or Level Up in skills and salary by completing SkillFactory paid online courses:
- Machine Learning Course a>(12 weeks)
- Learning Data Science from scratch (12 months)
- Analyst profession with any starting level (9 months)
- Python for Web Development Course (9 months)
- DevOps course (12 months)
- Profession Web developer (8 months)
- Trends in the Data Scene 2020
- Data Science has died. Long live Business Science
- Cool Data Scientists do not waste time on statistics
- How to become a Data Scientist without online courses
- 450 free courses from the Ivy League
- Data Science for the humanities: what is “data”
- Data Scenes on Steroids: Introducing Decision Intelligence