Skip to main content
. Author manuscript; available in PMC: 2022 May 10.
Published in final edited form as: Adv Neural Inf Process Syst. 2021 Dec;34:4974–4986.

Figure 7:

Figure 7:

IFM improves linear readout performance on all datasets for all ε ∈ {0.05, 0.1, 0.2} compared to SimCLR and MoCo-v2 baselines. Protocol uses 400 epochs of training with ResNet-50 backbone.