For BYOL, cosine similarity of two images should be evaluated by online_encoder or target_encoder?

lucidrains / byol-pytorch

Usable Implementation of "Bootstrap Your Own Latent" self-supervised learning, from Deepmind, in Pytorch

MIT License

1.76k stars 248 forks source link

For BYOL, cosine similarity of two images should be evaluated by online_encoder or target_encoder? #75

Closed vincezengqiang closed 2 years ago

vincezengqiang commented 2 years ago

cosine similarity of two images should be evaluated by online_encoder or target_encoder? if should be evaluated by online_encoder, should be evaluated before online_predictor or after online_predictor?

vincezengqiang commented 2 years ago

i find the answer in paper,