A Data Science Central Community
For this challenge, potential Facebook recruits will be exploring the map of the entire internet. Unlike the map of a city, where best routes are relatively fixed except for the occasional construction or parade detour, the paths that information travels over the web are constantly changing. There is no centralized system of stop-lights or traffic cops. Instead, there are tens of thousands of autonomous systems using a common protocol to advertise the next available hops, updated depending on service-agreements, capacity, and load. This will be a test of both the candidates engineering know-how and their ability to statistically learn on complex, dynamic graph structures.
The Task: you will be given a path which, at one point in the training time period, was an optimal path from node A to B. The question is then to make a probalistic prediction, for each of the 5 test graphs, whether the given path is STILL an optimal path. This is a much more difficult task than link prediction alone. The global structure of the graph may affect many optimal routes, paths can have varying lengths (and thus varying a priori probabilities of being optimal), and there may be multiple optimal routes for a given source & destination.
The Prize: Facebook is seeking data-savvy software engineers (Data Engineers) to build the next generation of systems that will transform the online experience of over a billion users. Appropriate candidates should have experience with multiple components across the big data stack (check out theVisualization track for another way to highlight your skills). There are many teams that they could be a fit for depending on their backgrounds . Positions are available in Menlo Park, Seattle, New York City, and London; candidates must have, or be eligible to obtain, authorization to work in the US or UK.
An example visualization of Internet topology (round-trip times) produced by Walrus
(courtesy of Young Hyun and inverted for display purposes)
Started: 4:02 pm, Wednesday 24 October 2012 UTC
Ends: 11:59 pm, Wednesday 21 November 2012 UTC (28 total days)
Read details at https://www.kaggle.com/c/facebook-ii