Speaker
Description
The MAIQMag project aims to utilize AI trained on multiple modalities to provide faster and more unique understanding of Hamiltonian parameters for Quantum Magnets. These modalities include, single crystal and powder inelastic neutron scattering data, Resonance Inelastic Xray scattering data and Magnetization. The neutron results are from Oak Ridge National Laboratory and the X-ray results are from Stanford Linear Accelerator, Argonne National Laboratory and Brookhaven National Laboratory. Furthermore simulations using the Su(n)ny.jl and EDRIX packages provide the necessary parameterized data for training. For this data to be accessed for training and interrogating an efficient data broker is needed. Tiled has proven to be an efficient solution for our use case. It provides a python scripting interface to allow quick searching of the metadata, and it allows extracting only the portions of multi-dimensional data that is needed for the processing. A hierarchical schema has been deployed and tests have been run as more data is ingested and have demonstrated query responses in a few ms. Currently there are 165,823 artifacts and 47,637 entities in the data base and it is ever growing.
