Recommender Systems
problem formalization
In the next video, I want to talk about recommendation systems. I want to talk about recommendation systems for two reasons:
First, simply because it is an important application in machine learning. Over the past few years, I have occasionally visited different technology companies in Silicon Valley, and I often chat with people working here who are working on machine learning applications. I often ask them, what is the most important application of machine learning, or, what are the most important applications of machine learning that you would like to improve? What are the applications of machine learning? The answer I hear most often is recommendation systems. Now, there are many groups in Silicon Valley trying to build good recommendation systems. So if you think about sites like Amazon, or Netflix, or eBay, or iTunes Genius , there are a lot of sites or systems trying to recommend new products to users. For example, Amazon recommends new books to you, Netflix tries to recommend new movies to you, and so on. These recommendation systems judge based on browsing what books you have bought in the past, or what movies you have reviewed in the past. These systems generate a large portion of the revenue for companies like Amazon and Netflix. Therefore, improving the performance of recommendation systems will have a substantial and direct impact on these enterprises.
Recommender systems are an interesting problem in academic machine learning. So we can go to an academic machine learning conference and the recommender system problem actually gets very little attention, or, at least, it's a very small share in academia. But if you look at what's going on, a lot of technology companies that have the ability to build these systems, they seem to be a high priority in a lot of companies. This is one of the reasons why I discuss it in this lesson.
The second reason I want to discuss recommender systems is that in the last few episodes of this class video I want to discuss some big ideas in machine learning and share them with you. We have also seen in this lesson that features are very important for machine learning. The features you choose will have a great impact on the performance of your learning algorithm.