Machine Learning Challenge
The task of the OCWC Challenge is to boost inter-connectivity between OCW training materials. Therefore, we plan to establish an infrastructure at five (or possibly more) selected OCW sites with an aim to create an inter-OCW site recommender system.
The technical task of the OCWC Challenge will require from each participant to create a recommending module in a form of a web service to generate recommendations generating as many as possible cross-OCW site end-user clicks. Challenge will provide a testing infrastructure allowing participants to access data about the OCW training materials and, in a form of a call-back web-service module, to propose which other links to show to a OCW web site user. Participating modules will be tested under AB-testing mode in real-time with real data from real users (i.e. each proposed module will get an equal share of the actual page views).
Modules will be ranked based on the number of clicks their recommendation generate and the top-ranked module will be declared the winner. The test will run for sufficient amount of time so that the differences in the final ranking will be statistically significant. Specific implementation details will be define in the following months.
There will be trail runs before the final test.