A convolutional neural network approach for reconstructing polarization information of photoelectric X-ray polarimeters

Kitaguchi, Takao; Black, K. M.; Enoto, Teruaki; Hayato, Asami; Hill, J. E.; Iwakiri, W.; Kaaret, Philip; Mizuno, T.; Tamagawa, Toru

doi:10.1016/j.nima.2019.162389

Cited by 21 publications

(10 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…(2) Combine predicted angles and their uncertainties in a weighted estimator of the polarization parameters p 0 , φ that maximizes the SNR. Our empirical findings indicate a substantial improvement over the current state of-the-art track reconstruction [3,11]. While the results shown here are specific to IXPE's GPDs, the methods are general, and can be applied to other imaging detector geometries.…”

Section: Introductionmentioning

confidence: 67%

“…In other words, θ follow the same distribution as θ but smeared by a modulation factor µ = I 1 (κ)/I 0 (κ). Current analyses [11,9] treat µ as constant for all tracks and calculate it based on broadband calibration measurements: no connection to individual track predictions. In effect, they assume homoskedastic emission angle measurements.…”

Section: Step Ii: Polarization Estimationmentioning

confidence: 99%

See 1 more Smart Citation

Deep ensemble analysis for Imaging X-ray Polarimetry

Peirson

Romani

Marshall

et al. 2021

Nuclear Instruments and Methods in Physics Research Section A:

View full text Add to dashboard Cite

X-ray polarimetry will soon open a new window on the high energy universe with the launch of NASA's Imaging X-ray Polarimetry Explorer (IXPE). Polarimeters are currently limited by their track reconstruction algorithms, which typically use linear estimators and do not consider individual event quality. We present a modern deep learning method for maximizing the sensitivity of X-ray telescopic observations with imaging polarimeters, with a focus on the gas pixel detectors (GPDs) to be flown on IXPE. We use a weighted maximum likelihood combination of predictions from a deep ensemble of ResNets, trained on Monte Carlo event simulations. We derive and apply the optimal event weighting for maximizing the polarization signal-to-noise ratio (SNR) in track reconstruction algorithms. For typical power-law source spectra, our method improves on the current state of the art, providing a ∼ 40% decrease in required exposure times for a given SNR.

show abstract

Section: Introductionmentioning

confidence: 67%

Section: Step Ii: Polarization Estimationmentioning

confidence: 99%

Deep ensemble analysis for Imaging X-ray Polarimetry

Peirson

Romani

Marshall

et al. 2021

Nuclear Instruments and Methods in Physics Research Section A:

View full text Add to dashboard Cite

show abstract

“…Especially at high energy, when the photoelectron track is longer, it is convenient to repeat the algorithm on the initial part of the track only, which is distinguished as the end with lower charge density. An alternative, which is particularly effective when the track has rich features, is to use more complicate techniques, such as the shortest path problem in graph theory [14] or convolutional neural networks [15,16].…”

Section: Reconstruction Of Photoelectron Trackmentioning

confidence: 99%

Analysis of the data from photoelectric gas polarimeters

Muleri¹

2022

Preprint

View full text Add to dashboard Cite

This chapter is dedicated to the description of the tools and procedures for the analysis of the data collected by X-ray photoelectric gas polarimeters, like the ones on-board the Imaging X-ray Polarimetry Explorer (IXPE). Although many of such tools are in principle common with polarimeters working at other energy bands, the peculiar characteristics and performance of these devices require a specific approach. We will start from the analysis of the raw data read-out from this kind of instruments, that is, the image of the track of the photoelectron. We will briefly present how such images are processed with highly-specialized algorithms to extract all the information collected by the instrument. These include energy, time of arrival and, possibly, absorption point of the photon, in addition to the initial direction of emission of the photoelectron. The last is the quantity relevant for polarimetry, and we will present different methods to obtain the polarization degree and angle from it. A simple method, used extensively especially during the development phase of X-ray photoelectric gas polarimeters, is based on the construction and fitting of the azimuthal distribution of the photoelectrons. While such a method provides in principle correct results, we will discuss that there are several reasons to prefer an analysis based on Stokes parameters, especially when one wants to analyze measurements of real, i.e., not laboratory, sources. These are quantities commonly used at all wavelengths because they are additive, and then operations like background subtraction or the application of calibration are trivial to apply, and they are normal and independent variables to a large extend. We will summarize how Stokes parameters can be used to adapt current spectroscopy software based on forward folding fitting to perform spectro-polarimetry. Moreover, we will derive how to properly associate the statistical uncertainty on a polarimetry measurement and the relation with another statistical indicator, which is in the minimum detectable polarization.

show abstract

“…Another independent and promising approach is a data-driven method, namely, a neural network. A deep neural network is capable of constructing a flexible model and is already applied to various observational techniques, such as X-ray polarimetry and gravitational wave detection [19,20,21]. A neural network model with a single hidden layer was proposed for reconstructing multiple Compton scattering events [22].…”

Section: Introductionmentioning

confidence: 99%

Event reconstruction of Compton telescopes using a multi-task neural network

Takashima,

Odaka,

Yoneda

et al. 2022

Preprint

View full text Add to dashboard Cite

We have developed a neural network model to perform event reconstruction of Compton telescopes. This model reconstructs events that consist of three or more interactions in a detector. It is essential for Compton telescopes to determine the time order of the gamma-ray interactions and whether the incident photon deposits all energy in a detector or it escapes from the detector. Our model simultaneously predicts these two essential factors using a multitask neural network with three hidden layers of fully connected nodes. For verification, we have conducted numerical experiments using Monte Carlo simulation, assuming a large-area Compton telescope using liquid argon to measure gamma rays with energies up to 3.0 MeV. The reconstruction model shows excellent performance of event reconstruction for multiple scattering events that consist of up to eight hits. The accuracies of hit order prediction are around 60 % while those of escape flags are higher than 70 % for up

show abstract

A convolutional neural network approach for reconstructing polarization information of photoelectric X-ray polarimeters

Abstract: a trade-off between the modulation factor and signal acceptance. The developed method with machine learning improves the polarization sensitivity by 10-20%, compared to that determined with the image moment method developed previously.

Cited by 21 publications

References 29 publications

Deep ensemble analysis for Imaging X-ray Polarimetry

Deep ensemble analysis for Imaging X-ray Polarimetry

Analysis of the data from photoelectric gas polarimeters

Event reconstruction of Compton telescopes using a multi-task neural network

Contact Info

Product

Resources

About