Fig. 3From: Sparse self-attention aggregation networks for neural sequence slice interpolationIllustration of how the parameter γ evolves during trainingBack to article page