Answers are grounded in this paper and cite their sources.
Bandit structured prediction describes a stochastic optimization framework where learning is performed from partial feedback.
No takes yet. Share an insight, caveat, or question.
Kreutzer et al. (2017) studied this question.