15-859(B) Course Project, a few more ideas

• Structured prediction: learning problems where the output is not just a +/- prediction or one of k values, but rather an object with more rich structure (a translation of a sentence, a tree, an image, etc).
• Inverse reinforcement learning: Given observations of people's behavior, can you find an objective function such that their behavior is consistent with optimizing this objective?
• Machine learning and auctions/pricing: What happens in auctions when bidders learn to adapt behavior from one auction to the next? Or on the other side, the use of learning by a seller to set prices to maximize revenue.