This post is a summary of various machine learning career advices by experts in this domain.
Introduction
The field of machine learning and artificial intelligence is very promising. The scientific progress and interest in the field are increasing every day. The applicability of the ideas into multiple fields makes it attractive to people from a diverse skill set.
Subject experts in this field are often transitioned from other technical domains like Physics, Economics, Social Sciences, etc. Although there are different roles under the umbrella of AI and ML, anyone who wants to enter the field faces the problem of information overload.
There are plenty of resources from which we can learn from. It is no wonder I a person is felt overwhelmed while transitioning into these fields. The best way to tackle this problem is to get advice from different people and try to make a judgment on which path might helpful to each individual.
The Solution
Over a couple of years, I have gone through multiple interviews where people have suggested their strategy to break into this field. In this post, I will be summarizing those, so that it will be easier for the next person.
The first on the list is a panel discussion that happened in the NYU AI School. I am providing the link to those talks as well.
The panel members were Sara Hooker, Pablo Samuel Castro, Chris Albon, and Marianne Monteiro. The talks start with each panel answering how they started into machine learning and the challenges they faced or are facing during the journey.
What is the best way to get into AI independently If I have a minimal background?
Narrowing down and getting the fundamentals right is important as far as the panelist is concerned. Marianne suggests the Coursera machine learning course as a good resource for fundamentals. Try to build things on your own and try to implement papers.
Pablo stresses the importance of learning linear algebra well for people wanting to get into deep learning. One another method he suggests is to reach out to internet communities where you can find people like you. I remember the meetup platform does a pretty good job of finding such communities where they organize virtual meetups/reading clubs etc.
The panelist says that it is going to take time and effort before you see results. Sara suggests starting with a book and sticking with it. The book she recommends is elements of statistics and the deep learning book (I think the Ian good fellow one). According to her, it is very important to have an applied project. Teach to learn.
Marianne says that it is natural to feel overwhelmed about the huge number of topics we are expected to know. She suggests working around projects and add to your repositories of knowledge every day.
Chris also says that he learned ML from books. One book he recommends is the introduction to statistical learning, elements of statistical learning. The way he learns is by creating atomic units of understanding from the learning. Sara says you have to be very good in one area and good enough in other areas. This reminds me of the ‘T’ shaped expertise discussed by Andrew Ng.
Chris says how concepts from different fields could be related and help in doing one better than doing by understanding it alone. I think he is mentioning something about multidisciplinary thinking where we distill ideas to the first principles.
One another advice Sara gives is that to try to be in a group of people with diverse sets. Work with people better than you and push you to make you better. Pablo says play to your strength and doesn’t bother much about your designation.
Sara says when you contact a researcher, you have to be specific about the topics. It is great if you can suggest some way you can improve their research work. You should contact them asking questions regarding their research work published. Pablo adds to this that you may try to replicate their work and improved results. This way you will have tangible results already and that will impress him a lot.
Conclusion
In essence, I am summarizing the following machine learning career advices from the talk.
- Create a project and share it with the community
- Learn the fundamentals right
- Try to learn and finish from the book by book
- Reach out to communities and be part of them
- Try to be very good in one area and good enough in other areas.
- Contact researchers with already tangible results
I will update with other similar events summarizing their pieces of advice as I encounter more. If you are looking for tutorial style articles related to machine learning and programming, you can find it here.