Figure 2
From: Estimating the effective fields of spin configurations using a deep learning technique
Training, validation, and testing of the network. (a) During the network training process, the training loss and validation loss values are calculated at each iteration. The black and red lines indicate the training loss and validation loss, respectively. (b) The estimated effective field (\({\varvec{F}}_{{{\varvec{x}},{\varvec{y}},\;{\mathbf{or}}\;{\varvec{z}}}}^{{*}}\)) from the samples’ test dataset is plotted as a function of the true effective field (\({\varvec{F}}_{{{\varvec{x}},{\varvec{y}},\;{\mathbf{or}}\;{\varvec{z}}}}\)). The black line shows \({\varvec{F}}_{{{\varvec{x}},{\varvec{y}},\;{\mathbf{or}}\;{\varvec{z}}}} = \varvec{F}_{{{\varvec{x}},{\varvec{y}},\;{\mathbf{or}}\;{\varvec{z}}}}^{{*}}\). The portion in the case of uniform magnetization is subtracted in the graph.
