Speakers
Description
We present an implementation of neural network based quantum state tomography following the framework of Koutný et al. (2022). The approach learns an inverse mapping from POVM measurement outcomes to the Cholesky parameters of a density matrix, ensuring positivity and physical consistency. Our work reproduces the original results for dimensionsd=2,5,7 using large scale synthetic datasets generated from Hilbert–Schmidt random states and square-root measurements. In addition to the replication, we extend the data-generation pipeline by introducing multiple noise models (including multinomial sampling noise and perturbations defined or called by the user) and packaging these tools into a modular Python library. We evaluate the network’s reconstruction accuracy under different noise levels and trial counts, highlighting the importance of consistent POVMs across training and testing. The resulting library provides a flexible platform for experimenting with neural-network tomography under realistic experimental conditions.