Learn Machine Learning 4X Faster by Participating in Competitions
Participating in competitions has taught me everything about machine learning and how It can help you learn multiple domains faster than online courses.
Image by author
You are stuck in a loop of learning new tools, new programming languages, and mathematics. In short, you have stopped progressing in your career. This is the slowest way to learn machine learning (ML), where we take multiple courses, learn mathematics, work on sample projects and eventually apply for a Job. It will cost you more than a year and eventually you will lose interest.
There is a better way of learning ML that involves understanding basics and then jumping right into solving machine learning challenges. In this blog, I am going to show you how I learned multiple domains (NLP, Computer Vision, ASR, RL, GAN) in machine learning just by participating in the competitions.
Image by Author | Kaggle
After learning the basics, I participated in the DrivenData Reboot: Box-Plots for Education challenge which taught me important skills of data manipulation, using logistic regression, and how to create SKlearn pipelines. It took me 12 days just to submit my first solution and after that, I used all my energy in improving the model performance metric which helped me reach 7th rank in the world. My second competition was on Natural Language Processing with Disaster Tweets | Kaggle and trust me I was bad at text classification but I put my time and energy into learning different techniques. After my second competition, I was over the moon of learning multiple domains, so I started participating in Zindi and other competitive platforms.
Highly reputed platforms
Check out ML Contests, if you are still unsure on where to start.
All competitive platforms come with a discussion where participants share the problem and try to figure out solutions together. This also means 3000+ teams are involved in solving a single problem. Collaborative learning also helps you understand what works for certain problems. You are also keeping up with the new trend and eventually, it’s preparing you for your future job as an ML engineer.
Image by Kaggle
You can also participate as a team of four and divide the workload. Working in a team also prepares you for real-life scenarios where the work will be divided among members within the data team. Most people say why can’t you Google a solution, but in my opinion, talking to your teammates and coming up with a solution is an effective way of solving problems.
Participating as a team also helps you build good communication skills that are critical if you are working remotely on projects. If you are new to machine learning your teammates can also help you learn new tools and skills.
Leaderboard and Prize
Leaderboard, ranks, medals, and prize money are perfect motivators for you to participate. This competitive environment also helps you push the boundaries and motivate you to work hard in achieving the top ten. When I started working on AI4D Baamtu Datamation — Automatic Speech Recognition in WOLOF — Zindi, I had no idea how to process the audio or how to train the model on GPU, but after 2 months of learning from blogs, YouTube, and GitHub projects I was able to reach 1st rank. This made me realize that I can learn anything. After this I have gained enough self-confidence to try new things and learn more tools. If I followed the ASR online course, it would have taken me four months to learn and another four months to get better at it.
To be honest you can win thousands of dollars just by working a little bit harder so why not make some money while learning new machine learning domains. To motivate you I am sharing the Kaggle competition prize pool.
My advice is to start small and keep working hard to reach the top rank. Reaching top will teach you important lessons that you will never learn from courses or tutorials.
Image from Zindi
Learning from an online course or getting a university degree won’t teach you about various machine learning problems and sometimes you don’t even have the skills to deal with unstructured data such as audio, video, or text. Either you will learn by taking multiple specialization courses which might take more than a year to learn or you can participate in various competitions with different domains.
Participating in challenges helped me understand problems that I never knew existed before and how to avoid model basis. It also motivated me to pursue my career in machine learning in production which is quite different from research based projects.
Machine learning consists of:
- Computer Vision
- Natural Language Processing
- Reinforcement Learning
- Time Series
- Tabular (Regression / Classification)
- Generative Learning
Image by magora-systems
These domains have sub-domains and it’s hard to learn everything about AI and various technologies, so it’s better to participate in current competitions that are solving current problems within an organization or company. You get to play around with production-ready data and learn various techniques to process the data.
If you want a smarter way of learning subdomain then participate in multiple competitions at once and don’t limit yourself with Kaggle, explore the competitions on other platforms that you like the most.
By participating in the competition we will learn to collaborate on a machine learning project, learn state-of-the-art techniques to tackle unknown problems, and learn various domains within the ML universe. Apart from learning you can gain recognition, glory and prize money.
We don’t need computing power, dataset, IDE ( environment), or knowledge to participate. Kaggle provides everything and we can learn unknown domains by reviewing other people’s solutions or asking questions on forums. The only thing that is stopping you to learn faster is your comfort zone and lack of awareness. I have seen many beginners start on Kaggle and now they are working for big companies such as NVIDIA, Ali baba, H2O, and Amazon.
In this blog, we have covered the importance of learning through competitions and how it can prepare you for the professional world. I hope you like it.
"There was a time when I was participating in 10 competitions on multiple platforms and those days were the happiest days of my life as I am a lifelong learner."
Abid Ali Awan (@1abidaliawan) is a certified data scientist professional who loves building machine learning models. Currently, he is focusing on content creation and writing technical blogs on machine learning and data science technologies. Abid holds a Master's degree in Technology Management and a bachelor's degree in Telecommunication Engineering. His vision is to build an AI product using a graph neural network for students struggling with mental illness.