Machine learning for X-ray astronomical spectroscopy

Overview

Development stage

In development or acquisitionSource: a) Pre-deployment – The use case is in a development or acquisition status.

High-impact designation

Not high-impactSource: c) Not high-impact

Impact justification

Not reported

Start date

Not reported

Withheld from public reporting

NoSource: a) No

Mission

Topic area

Science

Operational functions

No function tag matched Derived by keyword rules; see methodology

Problem the AI is intended to solve

We use Simulation Based Inference to construct 34000 artificial spectra that are representative of observed Active Galactic Nuclei X-ray spectra with NASA's NuSTAR X-ray telescope. Based on previous literature results for these spectra, we train, validate, and test a new neural network model, ML-mytorus, which predicts best values for key physical parameters fast, automatically, and with very high degree of accuracy, including error estimates. We make the code publicly available, and set up a dedicated webpage for the community to upload any spectrum and use the NN model. A publication is about to be submitted for review.

Expected benefits

Reproducibility: Standard spectral fitting is done interactively with human decision making. Automation and speed: Using the trained neural network model is completely automated, and takes just seconds for a given spectrum (compare to hours/days for interactive fitting). Openly available: Anyone can go to our webpage, load X-ray spectra in 2-column format and obtain results in seconds by clicking on a single button. Extensibility: The methodology can be extended for other X-ray (and non-X-ray) telescopes producing spectra, including for other types of astronomical observations, provided telescope and physical model characteristics are known. Training only needs a modest number (tens...) of existing previous fitted observed spectra, which are statistically perturbed to produce literally thousands of simulated spectra for training. Simulations, training and validation take no more than a few days with modest high-performance computing use.

System outputs

Predictions for astrophysical parameters with uncertainties

Technology

AI classification

Classical / predictive MLSource: Classical/Predictive Machine Learning: Models trained on data to make predictions or classifications based on identified patterns or relatio…

System name(s)

Not reported

Custom-developed code

Not reported

Public source code

Not reported

Sourcing

How it was built

Not reported

Vendor (as reported)

Not reported

Vendors (standardized)

None on the standard list

A vendor is the supplier named by the agency. It does not identify the underlying model or AI technology.

Data and privacy

Involves PII

Not reported

Privacy Impact Assessment

Not reported

Authorization to Operate

Not reported

Demographic features

Not reported

Training and evaluation data

Not reported

Federal Data Catalog

Not reported

Governance

0 of 8 minimum-practice questions answered. A blank answer means the agency reported nothing; it does not mean the practice is absent.

Potential impacts and how they were identified

Not reported