Free Book: Applied Stochastic Processes

If you look at chapter 5 (six degrees of separation) it applies to Youtube videos as well, in the sense that there is a path involving no more than six links from any Youtube video to any other one. Using a recursive algorithm for (automated) crawling is not a good idea though, as explained in chapter 5. Also, some videos are somewhat disconnected from the vast majority of Youtube videos. For instance, can you start with a video of the Beatles, and end up after any amount of browsing, discovering a machine learning video? Maybe not, and it means that the Youtube graph is not fully connected, and you need a number of seed videos from each connected component when doing your browsing, in order to retrieve all of them.