Astro arXiv | astro-ph.IM

Differentiable Cosmological Simulation with Adjoint Method by Yin Li et al. on Monday 21 November

Rapid advances in deep learning have brought not only myriad powerful neural

networks, but also breakthroughs that benefit established scientific research.

In particular, automatic differentiation (AD) tools and computational

accelerators like GPUs have facilitated forward modeling of the Universe with

differentiable simulations. Current differentiable cosmological simulations are

limited by memory, thus are subject to a trade-off between time and space/mass

resolution. They typically integrate for only tens of time steps, unlike the

standard non-differentiable simulations. We present a new approach free of such

constraints, using the adjoint method and reverse time integration. It enables

larger and more accurate forward modeling, and will improve gradient based

optimization and inference. We implement it in a particle-mesh (PM) $N$-body

library pmwd (particle-mesh with derivatives). Based on the powerful AD system

JAX, pmwd is fully differentiable, and is highly performant on GPUs.

arXiv: http://arxiv.org/abs/http://arxiv.org/abs/2211.09815v1

pmwd: A Differentiable Cosmological Particle-Mesh N -body Library by Yin Li et al. on Monday 21 November

The formation of the large-scale structure, the evolution and distribution of

galaxies, quasars, and dark matter on cosmological scales, requires numerical

simulations. Differentiable simulations provide gradients of the cosmological

parameters, that can accelerate the extraction of physical information from

statistical analyses of observational data. The deep learning revolution has

brought not only myriad powerful neural networks, but also breakthroughs

including automatic differentiation (AD) tools and computational accelerators

like GPUs, facilitating forward modeling of the Universe with differentiable

simulations. Because AD needs to save the whole forward evolution history to

backpropagate gradients, current differentiable cosmological simulations are

limited by memory. Using the adjoint method, with reverse time integration to

reconstruct the evolution history, we develop a differentiable cosmological

particle-mesh (PM) simulation library pmwd (particle-mesh with derivatives)

with a low memory cost. Based on the powerful AD library JAX, pmwd is fully

differentiable, and is highly performant on GPUs.

arXiv: http://arxiv.org/abs/http://arxiv.org/abs/2211.09958v1

H-FISTA: A hierarchical algorithm for phase retrieval with application to pulsar dynamic spectra by Stefan Osłowski et al. on Monday 21 November

A pulsar dynamic spectrum is an inline digital hologram of the interstellar

medium; it encodes information on the propagation paths by which signals have

travelled from source to telescope. To decode the hologram it is necessary to

"retrieve" the phases of the wavefield from intensity measurements, which

directly gauge only the field modulus, by imposing additional constraints on

the model. We present a new method for phase retrieval in the context of pulsar

spectroscopy. Our method makes use of the Fast Iterative Shrinkage Thresholding

Algorithm (FISTA) to obtain sparse models of the wavefield in a hierarchical

approach with progressively increasing depth. Once the tail of the noise

distribution is reached the hierarchy terminates with a final, unregularised

optimisation. The result is a fully dense model of the complex wavefield that

permits the discovery of faint signals by appropriate averaging. We illustrate

the performance of our method on synthetic test cases and on real data. Our

algorithm, which we call H-FISTA, is implemented in the Python programming

language and is freely available.

arXiv: http://arxiv.org/abs/http://arxiv.org/abs/2211.11129v1

3D Detection and Characterisation of ALMA Sources through Deep Learning by Michele Delli Veneri et al. on Monday 21 November

We present a Deep-Learning (DL) pipeline developed for the detection and

characterization of astronomical sources within simulated Atacama Large

Millimeter/submillimeter Array (ALMA) data cubes. The pipeline is composed of

six DL models: a Convolutional Autoencoder for source detection within the

spatial domain of the integrated data cubes, a Recurrent Neural Network (RNN)

for denoising and peak detection within the frequency domain, and four Residual

Neural Networks (ResNets) for source characterization. The combination of

spatial and frequency information improves completeness while decreasing

spurious signal detection. To train and test the pipeline, we developed a

simulation algorithm able to generate realistic ALMA observations, i.e. both

sky model and dirty cubes. The algorithm simulates always a central source

surrounded by fainter ones scattered within the cube. Some sources were

spatially superimposed in order to test the pipeline deblending capabilities.

The detection performances of the pipeline were compared to those of other

methods and significant improvements in performances were achieved. Source

morphologies are detected with subpixel accuracies obtaining mean residual

errors of $10^{-3}$ pixel ($0.1$ mas) and $10^{-1}$ mJy/beam on positions and

flux estimations, respectively. Projection angles and flux densities are also

recovered within $10\%$ of the true values for $80\%$ and $73\%$ of all sources

in the test set, respectively. While our pipeline is fine-tuned for ALMA data,

the technique is applicable to other interferometric observatories, as SKA,

LOFAR, VLBI, and VLTI.

arXiv: http://arxiv.org/abs/http://arxiv.org/abs/2211.11462v1

Development of a trigger for acoustic neutrino candidates in KM3NeT by Dídac D. Tortosa et al. on Monday 21 November

The KM3NeT Collaboration is constructing two large neutrino detectors in the

Mediterranean Sea: ARCA, located near Sicily and aiming at neutrino astronomy,

and ORCA located near Toulon and designed for the study of intrinsic neutrino

properties. The two detectors together will have hundreds of Detection Units

with Digital Optical Modules kept vertically by buoyancy forming a large 3D

optical array for detecting the Cherenkov light produced after the neutrino

interactions. To properly reconstruct the direction of the incoming neutrino,

the position of the DOMs, which are not static due to the sea currents, must be

monitored. For this purpose, the detector is equipped with an Acoustic

Positioning System, which is composed of fixed acoustic emitters on the sea

bottom, a hydrophone in each DU base, and a piezoceramic sensor in each DOM, as

acoustic receivers. This network of acoustic sensors can be used not only for

positioning, but also for acoustic monitoring studies such as bioacoustics,

ship noise monitoring, environmental noise control, and acoustic neutrinos

detection. This work explores the possibility of creating a trigger for saving

the data for ultra-high-energy neutrino candidates detected acoustically by the

hydrophones. The acoustic signal caused by the neutrino interaction in a fluid

is a short-time duration Bipolar Pulse extremely directive and with a Fourier

transform extending over a wide range of frequencies. A study of signal

detection, has been done by simulating BP produced by the interaction of a UHE

neutrino at 1 km from the detector at zero-degree incidence added to the

experimental real acoustic data. Finally, a trigger proposal has been developed

in order to record candidates of BPs and it has been tested. The number of

candidates per second, precision, and recall have been monitored according to

the cuts applied and parameters calculated by the algorithm.

arXiv: http://arxiv.org/abs/http://arxiv.org/abs/2211.10149v1

Transmission curves of narrow-band filters in large-FoV and fast astronomical instruments by Federico Battaini et al. on Monday 21 November

Narrow-band filters are often used to constrain the chemical composition of

astronomical objects through photometry. A challenge to derive accurate

photometry is that narrow-band filters are based on interference of multiple

reflections and refractions between thin layers of transparent dielectric

material. When the light rays reach the surface of a filter not perpendicular

to it, they cross the layers obliquely travelling a path longer than the

thickness of the layers and different for each inclination. This results in a

blue-shift of the central wavelength and a distortion of the transmission

curve. Hence, particular care should be taken when narrow band filters are used

in presence of small f-numbers and large non-telecentric angles, as frequent in

the large field of view (FoV) instruments. Sometimes, the broadening and

central wavelength shift of the transmission curve are considered and

compensated in the design of filters for instruments with a small f-number.

Here we consider the combined effect of small f-number, non-telecentricity and

large FoV. Where single spectral lines are considered, a shift in central

wavelength or a change in the shape of the transmission curve may introduce an

instrumental dispersion in luminosity and in the linked color indices. We found

that transmission curves of narrow band filters can be significantly different

in shape than the nominal ones. The bottom limits for filters' effective FWHM

for each f-number; the monotonic behavior of the blue-shift with distance from

the center of FoV; the monotonic quality decrease of the transmission curves

and the photometric dispersion introduced by the filters are computationally

estimated. This work could represent a useful tool to evaluate the fitness of a

particular filter at a particular facility.

arXiv: http://arxiv.org/abs/http://arxiv.org/abs/2211.10150v1