This course is all about A/B testing.
A/B testing is used in all places. Advertising and marketing, retail, newsfeeds, internet advertising, and extra.
A/B testing is all about evaluating issues.
For those who’re an information scientist, and also you wish to inform the remainder of the corporate, “brand A is best than brand B”, nicely you may’t simply say that with out proving it utilizing numbers and statistics.
Conventional A/B testing has been round for a very long time, and it’s filled with approximations and complicated definitions.
On this course, whereas we are going to do conventional A/B testing to be able to admire its complexity, what we are going to finally get to is the Bayesian machine studying method of doing issues.
First, we’ll see if we are able to enhance on conventional A/B testing with adaptive strategies. These all make it easier to remedy the explore-exploit dilemma.
You’ll be taught in regards to the epsilon-greedy algorithm, which you might have heard about within the context of reinforcement studying.
We’ll enhance upon the epsilon-greedy algorithm with the same algorithm known as UCB1.
Lastly, we’ll enhance on each of these by utilizing a completely Bayesian method.
Why is the Bayesian technique attention-grabbing to us in machine studying?
It’s a completely completely different mind-set about chance.
It’s a paradigm shift.
You’ll in all probability want to return again to this course a number of instances earlier than it totally sinks in.
It’s additionally highly effective, and plenty of machine studying specialists typically make statements about how they “subscribe to the Bayesian college of thought”.
In sum – it’s going to provide us a number of highly effective new instruments that we are able to use in machine studying.
The belongings you’ll be taught on this course usually are not solely relevant to A/B testing, however slightly, we’re utilizing A/B testing as a concrete instance of how Bayesian strategies will be utilized.
You’ll be taught these basic instruments of the Bayesian technique – via the instance of A/B testing – and then you definitely’ll be capable of carry these Bayesian strategies to extra superior machine studying fashions sooner or later.