How controllers from industrial machinery can coordinate multitask machine learning

Instead of compromising among parameter updates dictated by different training objectives, ControlG allocates computational capacity to objectives sequentially and dynamically.