Abstract: In this work, we optimize the scheduling of Deep Learning (DL) training jobs from the perspective of a Cloud Service Provider running a data center, which efficiently selects resources for ...