1 paper
Anna Heuser, Thomas Kesselheim
Motivated by stochastic optimization, we introduce the problem of learning from samples of contextual value distributions. A contextual value distribution can be understood as a fa…