Skip to main content
. Author manuscript; available in PMC: 2018 Jan 26.
Published in final edited form as: Adv Neural Inf Process Syst. 2017 Dec;30:3239–3249.

Figure 5.

Figure 5

(a) Learned TF frequency parameters for misspecified and normal TFs on MNIST. The mean field model correctly learns to avoid the misspecified TFs. (b) Larger sequence lengths lead to higher end model accuracy on CIFAR-10, while random performs best with shorter sequences, according to a sequence length calibration experiment.