Well, like most things of a certain level of sophistication, the answer is it depends. But before we delve into this matter, let’s start with defining what DS research is exactly. By this term, I refer to with the advancement of the field through the experimentation around new ideas, methods, techniques, and even the development and testing of new algorithms applicable to data science. Sounds like a lot but in practice there is a great deal of specialization so it’s not as overwhelming. For example, someone may do research in the data science technology focusing on distributed computing, while someone else focus on the design of a new supervised learning technique or a heuristic.
But don’t you need funding for all this? Well, in the traditional approach to research funding, usually in the form of grants sponsored by a government or some large organization, is something essential. After all, scientific research requires a great deal of resources and people who although passionate about the subject, may not work for free. Nevertheless, the expenses of research in data science are minimal, meaning that you don’t need a huge grant in order to get the ball rolling.
In essence, when you do DS research your key expenses are your time and the cloud computing rental. After all, Amazon and Microsoft need to make some money too when you are using their cloud services for your projects. Still, the prototyping is something you can do on your own computer so the cloud bill doesn’t have to be very high, unless you are working with a particularly large dataset, one that qualifies as big data.
I’m not saying that everyone can do data science research on his own. However, nowadays it’s easier than ever before to experiment without a lot of facilities or some sponsorship for a research project. People have been publishing papers on their own for years now and unless you want to do some large-scale research project, you can work with limited resources. And who knows, maybe this idea of yours can morph into a business product or service that can be a data science start-up. It doesn’t hurt to try!
A good tool for doing data science research is Julia, particularly through the Jupyter IDE. You can learn more about the language through the corresponding website, while my book on it can be a great resource for delving into it deeper. Note that the book was written for an earlier version of the language so the code may not be compatible with the latest version (v. 1.0) of Julia. Cheers!
Zacharias Voulgaris, PhD
Passionate data scientist with a foxy approach to technology, particularly related to A.I.