上 Adashare Learning What To Share For Efficient Deep Multi-task Learning 230921-Adashare Learning What To Share For Efficient Deep Multi-task Learning
2 Multitask learning is an open and challenging problem in computer vision The typical way of conducting multitask learning with deep neural networks is either through handcrafted schemes that share all initial layers and branch out at an adhoc point, or through separate taskspecific networks with an additional feature sharing/fusion mechanism Unlike existing methods, we This work proposes an adaptive sharing approach, called AdaShare, that decides what to share across which tasks for achieving the best recognition accuracy, while taking resource efficiency into account Multitask learning is an open and challenging problem in computer vision Adashare learning what to share for efficient deep multi-task learning