GRACE (https://t.co/KbrA7pHLlu) is a new framework for compressed communication in deep learning. If you train large models in a cluster of GPUs, compressed communication can speed-up training. Check: https://t.co/an7PQmVcbi
For anyone looking to exchange ideas on #AI & #MachineLearning, the @kaustMLhub is the perfect one-stop-shop. Prof. Bernard Ghanem launched the Hub with the first talk in a new seminar series: https://t.co/yRUW31rc3H