more data

Author: lucidrains

README.md

@@ -12,6 +12,8 @@ Update 1: There is now <a href="https://untitled-ai.github.io/understanding-self
 
 Update 2: A <a href="https://arxiv.org/abs/2010.10241">new paper</a> has successfully replaced batch norm with group norm + weight standardization, refuting that batch statistics are needed for BYOL to work
 
+Update 3: It <a href="https://arxiv.org/abs/2011.10566">turns out</a> even the momentum encoder is not necessary. You can have fully functioning self-supervised learning without negative pairs nor the exponential moving average of the target encoder. I'll update BYOL later with a flag that will allow you to turn off the EMA and break feature parity with 'SimSiam'.
+
 <a href="https://www.youtube.com/watch?v=YPfUiOMYOEE">Yannic Kilcher's excellent explanation</a>
 
 ## Install