Thursday 7th November, 2013
3:35pm to 4:10pm
Title: Real Time Processing of Graphical Models in Education
The speaker is Director of Data Science at Knewton, a digital education company that processes interactions streams from over 2 million students with compact content graphs in order to produce realtime recommendations. Knewton uses probabilistic graphical models to express the relational and temporal structure of each educational experience, inferring latent parameters of content, and of students, that explain observed interactions and inform our recommendations. We've encountered a number of interesting technical challenges while scaling our platform, including versioned content graphs, sparsity and temporal clustering in observed data, and wide variation in the key dimensions of our models. This talk will focus on the (sometimes) pragmatic approaches we've taken to engineering around those challenges.
Sign in to add slides, notes or videos to this session