What are some non-traditional approaches to learning machine learning?
So, these aren't quite courses or bootcamps, but I have learned a whole bunch from Microsoft Azure MLStudio and AWS's SageMaker. Both are free to use and they helped me grasp the concepts in a more concrete way.
I really, really liked the experiments in Microsoft Azure’s ML Studio. You get a whole bunch of free credits, you get to utilize Azure ML optimized cloud instances, and their experiment documentation is very good and easy to follow. If you are interested in checking out the experiments you can run, check here. The ML Studio has a webapp UI, but also supports Python/R. https://studio.azureml.net/
I also really like AWS’s SageMaker, which is built on Jupyter notebook. You get two months of SageMaker free with the AWS free tier, but you will definitely want to carefully read the documentation, as selecting the wrong instance type before training your model could result in racking up charges. https://aws.amazon.com/sagemaker/
I have two favorite online courses for deep learning - I do love this deeplearning.ai sequence on Coursera & I'd recommend starting with that. Afterwards, I also really like the fast.ai sequence. (I took followed the first half of fast.ai online, and then I applied to take the 2nd part in person at USF this past spring. They have an in person study group/open office hours every day which is amazing - both Jeremy Howard and Rachel Thomas are there most days.). It's also really important to try coding things on your own - either by extending & experimenting with one of the projects in a course, or by trying out a Kaggle competition.
I also like Berkeley CS 294 for Reinforcement Learning (though it requires a fair amount more math than either deeplearning.ai or fast.ai). The videos are available for free on YouTube.
This past summer I was part of the Scholars program at OpenAI, and I'd highly recommend it. It's a paid 3-month program where you're paired with a mentor and spend the time partly studying AI and partly working on an independent research project. They're starting another cohort in the winter of 2019, and they'll have applications available for that soon.
Hi @christine I applied for "Scholars program at OpenAI" this year but got rejected. Can you guide me how to prepare so that I can accepted next year?
OpenAI is planning to have another Scholars program starting in early 2019, so the applications should be quite soon. Last time they had many, many applications - I think it was around 600 or 700, and so I'm sure they must end up missing a lot of very qualified people.
I don't have any insider knowledge about how the applications are reviewed, but here is my perspective from having gone through the process as an applicant. My guess is that they are looking for candidates who have excelled in some other field, who bring a unique perspective or background, and who show that they are self-motivated and would do well on an independent project. The scholars program is very self-driven: you plan your own syllabus to study, and you do your own independent final project. You meet with a mentor each week, which is very helpful, but you set your own daily schedule and goals. It's important to show that you would be self-motivated and would thrive in this kind of environment.
In preparing for applications (both for this program and for others), I worked through this deeplearning.ai sequence and became a mentor for the final two courses. I also worked through the fast.ai sequence, and then I put a project up on github that I felt represented my skills. (Advice I got elsewhere about github: it's better to have one really fantastic project, rather than several half-done ones. In particular, it's good to have a nice webpage or presentation for that project, so that people are immediately impressed.)
Lastly, I'd say - try to emphasize what makes you unique, and tell why you're excited about AI. It's a small group of people who read the 600+ applications, so you'll want to have something that makes you stand out as an individual.
Good luck with the process! Feel free to reach out with any other questions.