I'd be interested in knowing how much deep learning is changing the algorithms used in this field, given the performance of restricted boltzmann machines on the netflix data set http://www.cs.utoronto.ca/~hinton/absps/netflixICML.pdf.
You'd be surprised how overwhelmingly common and effective very simple methods like logistic regression and basic decision trees are for such systems.
Further, RBMs and other deep learning tools require a significantly more sophisticated mathematical background than algebra and a much broader understanding overall.
The netflix prize touched on one of many areas related to recommender systems.
As mentioned already, very simple methods can be really effective. Things such as the UI are also known to have a big impact on how 'useful' people find the recs.
I am surprised that there is no mention of this in the course syllabus -- in fact it looks like a lot of recent techniques that are missing. They don't even talking about LSA(/SVD)-based methods until the end of the course.
Further, RBMs and other deep learning tools require a significantly more sophisticated mathematical background than algebra and a much broader understanding overall.