Fig. 7: Trace-preserving regularization.
From: Quantum process tomography with unsupervised learning and tensor networks
a Tensor network for Δϑ, obtained by subtracting the identity MPO (with bond dimension 1) to the (properly normalized) LPDO Λϑ, and tensor contraction required to compute Γϑ. We show the measurement of the regularization Γϑ (b) and the reconstruction infidelity (c) during the training for a one-dimension random quantum circuit with N = 10 qubits and depth D = 2, for different number of the total data set size M.
