Posts

Showing posts from February, 2022

Sparse Multilingual NMT

 The idea is to use a sprase and big architecture for multilingual NMT. This will mirror the pathway architecture for large-scale evolving multitask learning systems.