part0003

Hello! If you’re reading this, you’ve purchased this course on recommender systems either in PDF or book form. Thank you! If you’re also interested in the video version of it, you’ll find it at www.sundog-education.com.

I’m Frank Kane, CEO of Sundog Education. I spent nine years at Amazon.com, where I dedicated most of my career to building various parts of their recommendation systems and managing teams responsible for them. You know “people who bought this also bought?” Yeah, I ran that for awhile, along with their “personalization platform” team. I have a lot of real-word experience to share with you, combined with the latest research I’ve done on new developments in the field since I left Amazon. Recommendations are one of the most fascinating applications of machine learning, and also one of the most lucrative – recommending products is central to Amazon’s success, recommending movies is central to YouTube and Netflix, and you can even think of Google as just recommending web pages and ads to people.

This isn’t going to read like a typical book. What you have here are the slides I’ve used for presenting this information in video or live form, along with the script I prepared to accompany these slides. So for each topic, you’ll see an image of the slide associated with it, along with the text I wrote to explain each slide. But, it works surprisingly well – it has all the textual information you’d get in a typical book, but with many more visual aids, and a more casual, conversational tone than you’d find in a textbook. It’s ideal for visual learners who just find reading material a lot more efficient than listening to someone read it aloud in a video. I’ve written this script with this written version of the course in mind, and have attempted to make sure everything works just as well in print as it does in a video presentation.

You’ll find that code walk-throughs work a little bit differently in this format than what you may be used to from other technical books. Early in this course, you’ll be directed to download all of the code that accompanies it. When we get to slides that review this code, you’ll want to pull up that code on your computer as directed in those lectures, and refer to it alongside these written notes that explain what each section of the code does. Really, there’s no other practical way to do it – recommender systems involve a fair amount of code, and in most cases it simply won’t fit on one written page. I promise to be very specific about what parts of the code I’m talking about as we go through it, so you won’t get lost.

Oh, and if Amazon’s lawyers are reading this – don’t worry. I’ve been careful to only cover algorithms and techniques that have appeared publicly, in print. I’m not revealing any inside, confidential information here – although most of what we did at Amazon has been published at this point, anyhow .