2017.03.01 - Fast.ai - How to change careers and become a data scientist - one quant's experience
2017.04.06 - Fast.ai - Alternatives to a Degree to Prove Yourself in Deep Learning
I think I got this from Sam Altman recommending it on Twitter. It looks pretty interesting.
TODO: Summarize this.
Quora - What distinguishes reinforcement learning from deep learning?
My favorite answer:
Deep learning is an approach to implementing function approximation.
Reinforcement Learning is a learning problem in which the goal is to learn from interaction how to act in an environment to maximize a reward signal. There are many different algorithms to solve RL, many which involve function approximation, which when coupled with the above is what leads to deep reinforcement learning.
Amazon - Reinforcement Learning: An Introduction
Highly rec'd book
2017.10.28 - Medium - David Foster - AlphaGo Zero Explained In One Diagram
2018.01.26 - Medium - David Foster - How to build your own AlphaZero AI using Python and Keras
2017.05.25 - MonkeyLearn.com - A practical explanation of a Naive Bayes classifier
In simple terms, a Naive Bayes classifier assumes that the presence of a particular feature in a class is unrelated to the presence of any other feature. For example, a fruit may be considered to be an apple if it is red, round, and about 3 inches in diameter. Even if these features depend on each other or upon the existence of the other features, all of these properties independently contribute to the probability that this fruit is an apple and that is why it is known as ‘Naive’.
http://ruder.io/optimizing-gradient-descent/
This is apparently a "plain English" explanation of different gradient descent algorithms. I found out about it from a tweet that was retweeted by Jeremy Howard.
2017.07.28 - Mashable - This high school kid taught himself to be an AI wizard
incredibly, Mikel's programming skills in machine learning and AI are almost completely self-taught — there are no courses on building AI-powered systems at his high school in Guildford, just outside of London. He spent big chunks of the last couple of years researching AI and machine learning on the internet, reading articles and watching videos.
"There are lots of free courses online, but I actually didn't take courses. When I had a big problem I wanted to solve, I would just Google around and try to find out about it, so I didn't follow any predefined track. I read about Kaggle online, and I thought, 'Why not try it?'"
After entering various competitions over several months and slowly placing higher and higher, Mikel helped create an algorithm for using computer vision to analyze 8 million YouTube videos to create accurate tags. His team ended up 7th out of 650 teams.
2017.10.24 - Rainforest QA - Kaggle Perks in Real Data Science Work
He rec's the following articles / forum posts:
For non-trivial validation:
Time-series validation when you need to predict for some specific time in the future
TODO: Grab and organize the rest of the links in this post.
https://www.kaggle.com/c/planet-understanding-the-amazon-from-space/discussion/36809
This is the one Justin went to.
2015.10.05 - Nicole White - Improving My CLI’s Autocomplete with Markov Chains
This is very easy-to-follow; it would be easy to use the code here as a template for another project.
Tutorials
This guy seems really smart, I'd really like to work with him.
Talks
Erik Mueller (he was on the IBM Watson team)
Apparently a good overview of all of the ML stuff out there.
Programming Collective Intelligence
Rec'd by someone who commented on another person's review for 'The Master Algorithm'.
rec'd by Bob
This is a wrapper for Amazon Mechanical Turk, useful for labelling data