This review serves as a concise introductory survey of modern compressive tomography developed since 2019. These are schemes meant for characterizing arbitrary low-rank quantum objects, be it an unknown state, a process or detector, using minimal measuring resources (hence compressive) without any a priori assumptions (rank, sparsity, eigenbasis, etc.) about the quantum object. This paper contains a reasonable amount of technical details for the quantum-information community to start applying the methods discussed here. To facilitate the understanding of formulation logic and physics of compressive tomography, the theoretical concepts and important numerical results (both new and cross-referenced) shall be presented in a pedagogical manner.
Quantum tomography is a class of characterization methods frequently used in current experiments, but its standard protocols suffer from unreliability originated from preknowledge assumptions. Self-consistent quantum tomography is an approach to avoid the problem, which treats every quantum operation in a characterization experiment as unknown objects to be characterized. As compensation for the beneficence, it leads to a problem that its characterization results cannot be determined uniquely only from experimental data due to the existence of experimentally undetectable gauge degrees of freedom, and we need to introduce a criterion to fix the gauge. Here, we propose to use a regularization technique to fix the gauge. First, we derive a sufficient condition on a characterization experiment to obtain all information of objects to be characterized except for the gauge. Second, we propose a self-consistent data-processing method with regularization and physicality constraints. A careless use of regularization can lead to non-negligible bias on the characterization result. As a solution for the concern, we propose a concrete way to tune the strength of the regularization, and mathematically prove that the method provides characterization results that converge to the gauge-equivalence class of the quantum operations of interest at the limit of data going to infinity. The asymptotic convergence guarantees the reliability of the method. We also derive the asymptotic convergence rate, which would be optimal. These theoretical results hold for any finite-dimensional quantum systems. Finally, as its first numerical implementation, we show numerical results on one-qubit system, which confirm the theoretical results and prove that the method proposed is practical.