Making ML Model Error an Information Source

This talk demonstrates how machine-learning residuals can reveal information about the system being observed. A mobile-network example uses them to assess how configuration changes affect performance indicators.