HIGHLIGHTS
- who: TYPE and collaborators from the Tsinghua University, China have published the research work: XMA : A crossbar-aware multi-task adaption framework via 2-tier masks, in the Journal: (JOURNAL)
- what: The main reason is that these methods completely freeze the weights of the backbone model and only apply the binary (Mallya et_al, 2018; Zhang et_al, 2022a) or shift-value (Zhang et_al, 2022b) masks, causing limited optimization space for learning new tasks. To tackle these issues, in this work, the authors propose XMA2, a novel ReRAM crossbar-aware learning framework via 2-tier masks for . . .
If you want to have access to all the content you need to log in!
Thanks :)
If you don't have an account, you can create one here.