1. Don't Go after Hype:
Machine Learning is super over-hyped by media. Yes, it is the future but don’t just run after the hype. First, take a step back and see if this is the thing that really interests you. You should experiment with it for a month or so, and if you are genuinely interested then you are good to proceed further. There are ocean of things and complex concepts to learn and you would end up getting demotivated if you are not interested in it. If you are not interested, then it is not the end of the world, there are infinite opportunities in this world. Find something that interests you, and work upon that.
2. You can’t run away from Mathematics:
Machine Learning is fully based upon mathematics. Some people say that it is not important and there are many libraries and frameworks by which you can implement algorithms. But trust me this is not the right way to learn. Try to think in this way, if you can just call some functions from these libraries and do the job, then why would companies hire people and pay them thousands of dollars. You must know how the things are working under the hood. And if you are someone, who was not interested in Mathematics in his/her high school days, try learning the maths behind the algorithms, you will fall in love with it in no time. Because you can actually see the results of the cryptic equations spreading their magic in real-world scenario. This will motivate you to learn more and deep-dive into the maths of Machine Learning.
3. Python or R?:
Many newbies are confused between these two languages. My personal advice would be to learn Python if you want to go into Deep Learning and R if you want to just perform some statistical analysis on the data. R lacks good packages when you go to Deep Learning whereas Python really shines there. But it is not that you should not learn R. Some of the job requirements may require you to use R at work. My advice would be, first get comfortable with Python and then you can pick up R in just a week when you feel like learning it. It’s good to have both of the languages in your arsenal.
4. Don’t run after certifications:
Many people just do the course for the sake of getting a certificate. But, turns out that this certificate hardly adds any value to your resume. The thing which really adds value is that what you have learned from that course, and how you have implemented these learnings in your own projects. There are many great courses available but you shouldn’t do them just for getting certificates, instead you should use them to learn and strengthen your concepts and apply them to solve real-world problems.
5. Machine Learning is not just about building models:
Many of the beginners think that we have to just do .fit() (this came from sci-kit learn 😁) and voila the model will be created. But this is not the case. You have to go through the complete life cycle of a data science project. You first have to define what problem you are aiming to solve. Then you must collect the relevant data, do exploratory data analysis on the data and understand it, handle missing values, do feature engineering, feature selection, try out different models, tune the hyperparameters, take your model into production, explain your project to the clients and how it may impact their business and there are lot more things to do. I will cover this more in depth in my following posts.
6. Get out of tutorial hell
Some people get stuck in the tutorial hell. Tutorial hell means that people get stuck on watching more and more courses. Watching courses are good but till a certain extent. They are just to provide you foundational knowledge. You must work on a dataset or a project after you build some good foundational knowledge. Trust me, you will learn most of the things by doing projects, working on a dataset on your own and taking a real-world scenario and applying machine learning to it. This will give you immense knowledge which no course can give you.
7. Do some really good projects to stand out
This is a follow up to the previous point, you should do some really good projects. Do not take on complex projects initially when you have just started out, try to do some beginner level projects. Then, once you get comfortable, you can do some really advanced project.
You should do atleast some projects totally from scratch. From scratch means, you should first define a problem statement .Then start collecting the data yourself (not use the ready ones),clean the data, handle missing values, perform feature engineering, feature selection, modelling, inference, deploy into an application and write a paper about it. It will give a great learning experience and would allow you to understand how things really work in industries.
8. Do not let your models die in Jupyter notebooks:
Even if you perform every step in a project and build a successful model, don’t let it die in the jupyter notebook. Your model is of no use until you take it into production. Try to develop a web application or an android application and deploy it to the cloud and see how the model performs in the real-world. Try to find the caveats into the system and improve upon them.
9. Share your learnings
You should give back to the data science community by sharing your knowledge. If you have made a project, write a blog post about it and explain others about it. If you want to share your knowledge or teach, maybe you can start a YouTube channel. This will help you to create an audience and also improve your story-telling and communication skills which are very crucial in this field. Lastly, make good connections on LinkedIn.
10. Don’t give up and learn everyday
My last advice to you all is to not give up. Things may get harder, but you have to stand firm and face failures and rejections.
Learn everyday. Consistency beats intensity. Just devote some time every day and you will start seeing results.
So these are my 10 tips for anyone who wants to start or is learning Machine Learning. There are much more tips I want to share, but let’s close it here.
Thanks for reading :)