For industrial-scale advertising systems, prediction of ad click-through...
The problem of attributing a deep network's prediction to its
input/base...
Data cube materialization is a classical database operator introduced in...
Gradients have been used to quantify feature importance in machine learn...