Shi-Shang Wang

Ordered Local Momentum for Asynchronous Distributed Learning Under Arbitrary Delays

Momentum SGD (MSGD) serves as a foundational optimizer in training deep models due to momentum