search

MishaLaskin / curl

CURL: Contrastive Unsupervised Representation Learning for Sample-Efficient Reinforcement Learning
MIT License
561 stars 88 forks source link

What is "ema"? #18

Closed slerman12 closed 3 years ago

slerman12 commented 3 years ago

In CURL.encode, what is the arg "ema"?

MishaLaskin commented 3 years ago

exponentially moving average of the key encoder

On Fri, Dec 11, 2020 at 5:57 PM Sam Lerman notifications@github.com wrote:

In CURL.encode, what is the arg "ema"?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/MishaLaskin/curl/issues/18, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABHWQWJDCIEIYQGNFJB3I5DSUKPWDANCNFSM4UXMCHRA .

slerman12 commented 3 years ago

It just occurred to me that query and key have different encoders. Is there a reason for this?

MishaLaskin commented 3 years ago

Yes - referring you to MoCo https://arxiv.org/abs/1911.05722 or you could read the implementation section of the CURL paper

On Mon, Dec 14, 2020 at 1:01 PM Sam Lerman notifications@github.com wrote:

It just occurred to me that query and key have different encoders. Is there a reason for this?

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/MishaLaskin/curl/issues/18#issuecomment-744610396, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABHWQWKELOBM2C6ODXLGJWTSUZHF5ANCNFSM4UXMCHRA .