Figure 5
From: Estimating the effective fields of spin configurations using a deep learning technique
Data generation of the trained network, MC, and greedy algorithm. (a) Spin configuration data are generated during 50 iterations by feeding a random map to the trained network. (b) Another random map is fed to the trained network and shows the result of 50 iterations. (c, d) The results show the data generated after 5000 iterations using the MC method and greedy algorithm. Each Hamiltonian parameter used the same data used for training. (e) Energy changes during iterations are shown for each method. The black and red data points represent data generation from different random maps of the trained network. The green and blue points indicate the MC method and greedy algorithm.
