For the first time in my life, choosing courses is not a constraint-satisfaction-problem. Coursework in grad school is supposed to be a structured approach to filling gaps in your knowledge so you can do better research. I was faced with a choice between intermediate-stats (a first-sem stats course - although, LW's blog says it has been tweaked so that computer-scientists continue to find it sexy) and ML. I actually picked up a decent enough amount (I think) of learning theory and topics like VC Dimension, Model selection etc. from the Vishy/Smola textbook and Vishy's course @ Purdue.
Anyway, it is ML and NLP Algos for this semester - that will be my coursework.
Next of course, is the topic I am freaking out about - LTI IC.
We listen to talks and try to find an advisor. I will report back about how well/poorly this goes :_(.