I just finished my master degree in engineering in France and I am now doing a double degree in data analytics at Boston University. I will be graduated in September and after that I would really like to do a PhD program in Machine Learning / Deep Learning but since my school in France and college in Boston are not really implicated in Deep Learning research I am struggling a bit to find information.
For 2 years now I have been passionated for Machine Learning, I have done some research for organizations using Deep Learning for Computer Vision and different Machine Learning for forecasting on numeric datas, a few kaggle challenges and some MOOCS ( FastAI, Machine learning and Deep Learning Specialization from Andrew Ng and Computer Vision from Aaron Bobbick ). However, I want to learn more, and I think doing a PhD in deep learning could really help me gain some knowledge. I am not afraid to do something new and I could even take a MOOC along with classes before starting the program.
Do you agree with my vision of the PhD or am I missing something ?
What are for you the most promising/ interesting PhD subjects in Deep Learning now ?
What advice would you give me to find a PhD program ? ( I could do it in USA but also in other countries like France or Canada )
Thank you so much
Btw : If you have a machine learning or computer vision project that you want to share send me a message, I love to talk about that kind of stuff and we could maybe do a working group
Do you want to study Deep Learning methods on a more theoretical level or do more practical applications? I’m doing a PhD in Geosciences and I’m studying Deep Learning by myself and trying to apply it to several problems in the topic of my thesis. It is being great! But I wouldn’t probably do a PhD in Deep Learning since my focus is more the practical applications. For me kaggle is being the perfect place to learn
This blog post by Rachel may also be of interest for you: https://www.fast.ai/2018/08/27/grad-school/
Thank you vey much for your answer, I read the article, it was really interesting and one of the first time I saw this point of view on PhD.
This is actually part of my question, I have a lot of ideas for practical applications that could work using Deep learning but I have a feeling that focusing on theorical level could be more interesting for me because I am afraid that if I used a technology during my PhD on a practical application, it will be obsolete by the end of my PhD. As stated in the article, specializing in a narrow field can be a trap but I also see it as the opportunity to have a rare profile because you master some competencies that just a few people do.
As computer science is my major I am more looking to do research that could apply to a range of projects than apply directly Deep Learning on a particular project
If you think on the % of people in the world you can build and train a Deep Learning model you have a rare profile anyway. My concern is that the DL field is growing so fast that a 4 year PhD program may be too long to dedicate to a specific problem.
Another option that seems good is OpenAI for example (https://openai.com/jobs/), you could try to apply for a Machine Learning Engineer position. As they say “the problems we pursue are always at the edge of the state of the art, which means that nobody’s an expert”.
Hi Etienne, going through a PhD is a very formative experience especially in Europe. Of course it is hard and challenge and it will cost you a lot of time, but at the end you will be able to see things differently.
Tech and academia are complementary, but keep in mind that research in academia should be driven by knowledge for it’s own sake, while in tech is only driven by profit(look at what happened at DeepMind recently).
Deep learning hype is now and nobody knows what else is coming tomorrow. But being in academia, and in a good group, will let you be in touch with the smartest people in the world without signing any NDA and see new trends in their early state. My advice is simple : give it a try!
That exactly how I see this opportunity, I really like the idea to work in order to develop competencies in a field and I think I need a little more time to experiment Deep Learning principles, which are quite new to me, before diving in the professional world. That’s why I am trying to find a PhD program in order to have the opportunity to work on a set of subjects and gain knowledge.
Thank you very much, do you have an idea in which academia it would be the best to attend in order to do a PhD program ?
Actually applying to OpenAI is more like a dream goal than a first step, I would be really happy to work there but I don’t think I have the level to. Even if I worked on a few Deep Learning projects they seems to be incredibly strong. I would surely try to apply after a PhD or a few years of practice
Happy we share this point of view! You should start looking at “how to write/plan a phd proposal”, just to have an idea since it is easier to get a job in tech than getting accepted in a phd program especially within a great group. About your question I can only suggest places in the DL field that are at the top since I have no idea about small groups :
The first is totally deep learning(actually dl was invented there), CMU is the greatest place for machine learning, and finally, Oxford could be my favorite choice as it is one of the top 5 universities in the world and it is an incredible place but wait to see what Brexit will be.