Tensor-network discriminator architecture for classification of quantum data on quantum computers

We demonstrate the use of matrix product state (MPS) models for discriminating quantum data on computers using holographic algorithms, focusing classifying a translationally invariant based $L$ qubits extracted from it. detail process in which single-shot experimental measurements are used to optimize an isometric tensor network, tensors compiled into unitary operations greedy compilation heuristics, parameter optimization resulting circuit model removes post-selection requirements model, and is inferenced either or entangled data. our training inference architecture synthetic dataset six-site bulk one-dimensional transverse field Ising (TFIM) deep its antiferromagnetic paramagnetic phases. experimentally evaluate Quantinuum's H1-2 trapped ion computer input modeled as invariant, bond dimension 4 MPSs across known phase transition TFIM. Using linear regression near point, we find predictions critical $h=0.962$ $0.994$ network discriminators $\chi=2$ $\chi=4$, respectively. These compare favorably with location $h=1$ despite far point. Our techniques identify families short-depth variational circuits data-driven hardware-aware fashion robust classical precondition parameters, can be adapted beyond machine learning myriad applications networks computers, such simulation error correction.