Figure 2
From: Accelerated spin dynamics using deep learning corrections
Spin-spin correlation using reference trajectory generated at \({\uptau =10^{-6}}\). Analysis of the mean of correlation \(\mu _{\xi (t)}\) as a function of time on \(4 \times 4 \times 4\) cubic lattice at \(\mathbf{a} , k_{B}T/J=0.4\), \(\mathbf{b} , k_{B}T_{c}/J\approx 1.44\), and \(\mathbf{c} , k_{B}T/J=2.4\) and those on \(8 \times 8 \times 8\) cubic lattice at \(\mathbf{d} , k_{B}T/J=0.4\), \(\mathbf{e} , k_{B}T_{c}/J\approx 1.44\), and \(\mathbf{f} , k_{B}T/J=2.4\). Blue line presents the Deep Learning (DL) result while black line, yellow line, and red line are the simulation results for \(\uptau =10^{-1}\), \(\uptau =10^{-2}\), and \(\uptau =10^{-3}\), respectively. Especially, at \(k_{B}T/J=1.44\) and \(k_{B}T/J=2.4\), green line and violet line show the simulation results for \(\uptau =10^{-4}\) and \(\uptau =10^{-5}\), respectively. g Threshold time \(t_{thres}\) as function of temperature. Filled rhombi (
) represents the Deep Learning result while filled black triangles(\(\blacktriangle\)), filled yellow circles (
), filled red squares (
), filled green inverted triangles (
), and filled violet pentagons (
) are the simulation results without DL corrections for \(\uptau =10^{-1}\), \(\uptau =10^{-2}\), \(\uptau =10^{-3}\), \(\uptau =10^{-4}\), and \(\uptau =10^{-5}\), respectively.
