Fig. 7: Partial training.
From: Neural network approach to quasiparticle dispersions in doped antiferromagnets
Separate amplitude (dark blue) and phase (light blue) training, for ground states of the t − J model on a 4 × 4 square lattice with t/J = 3, open boundaries and without holes Nh = 0 (a) and with a single hole Nh = 1 (b), compared to the full training in red. Thick lines and markers represent averages over 100 epochs (each with 200 samples) to guide the eye, with the respective standard deviation denoted by the error bars. We use a hidden dimension of hd = 70.
