Since forming the IDAO, I have had the opportunity to expand my own analytical skills to include skills in machine learning. As with the statistical learning, this knowledge has come largely through reading several excellent textbooks (see below), as well as use of online resources. Unlike the application of many purely statistical approaches, implementation of machine learning also requires an understanding of computer science and data organization. Much of my experience in these realms came out of necessity, and thus through trial and error rather than through study of a formal reference. However, many of these references include these methods to various degrees as well as the fun stuff. To see how to locate each source, click on the textbook or picture.
This very manageable textbook is a perfect starting point for anyone interested in learning about machine learning. In addition to a very nice overview of many methods of 'shallow' learning, it also includes R code to run the models yourself.
This somewhat larger and more technical textbook is the more advanced version of Introduction to Statistical Learning, by the same authors. It does not include code references, and remains at the 'shallow' machine learning level, but nonetheless provides a very nice comprehensive overview of these approaches.
Although I'm not sure HMM's are technically considered 'machine learning', I would strongly recommend this book, which I've mentioned in prior posts, if you plan to delve into deep learning approaches with recurrent neural networks or reinforcement learning
Before you dive into any online sources, blogs, or other textbooks, I'd highly recommend starting with this very basic, but understandable textbook. Not only does it provide an outstanding introduction to concepts like stochastic gradient descent, but it also teaches you to use to use Jupyter notebook and Python in a step-by-step manner.
Neural Networks and Deep Learning. This online textbook is definitely one of the best places to start for understanding deep learning, and a free resource at that.
Outstanding resource for using sklearn for 'shallow' learning approaches with code to run in Python. The description of deep learning is outstanding, although much of the code did not work for me and eventually I resorted to using Keras to run the models with Tensorflow.
There are numerous online videos and posts that I used to supplement my knowledge. The following were a few that I went back to several times:
The videos by Hugo LaRochelle (First have to find the ones in English) were an amazing compilation of information about deep learning.
This video by Brandon Rohrer is another great explanation of the challenging topic of Long-short-term memory and recurrent neural networks.
These deep learning videos were also well done, and easy to follow along.
There are undoubtedly other resources that I used online, not to mention the old standard Wikipedia, and like most things these days, when I got stuck or needed additional explanation, I simply googled whatever part I needed more help understanding.
Mike Rosenberg
March 29, 2018