SKACH Spring Meeting

1 Jun 2023, 09:00 → 2 Jun 2023, 17:30 Europe/Zurich

Aula (University of Geneva, Geneva Observatory)

Joanna Jermini-Howard, Carolyn Crichton

We are pleased to invite you to participate to the SKACH Spring Meeting hosted by our collaborators at the University of Geneva. This event will focus on consortium-wide updates, and updates from the five SKACH programs: Comms & Outreach, Instrumentation, Computing Platforms & Infrastructure, Data Science & Simulations and Science (AstroSignals).

SKACH is primarily funded by the State Secretariat for Education, Research and Innovation (SERI), and is responsible for managing the Swiss contribution to the SKAO. SKACH also leverages funding from external grants such as AstroSignals SINERGIA, SNF Bilateral, PASC, InnoSuisse, and Horizon Europe.

This endeavor is lead by a strong contingent of Swiss institutions; including, Universität Basel (UniBas), Fachhochschule Nordwestschweiz (FHNW), Universität Zürich (UZH), Eidgenössische Technische Hochschule Zürich (ETHZ), École Polytechnique Fédérale de Lausanne (EPFL), Zürcher Hochschule für Angewandte Wissenschaften (ZHAW), Université de Genève (UniGE), Haute École spécialisée de Suisse Occidentale (HES-SO), Centro Svizzero di Calcolo Scientifico (CSCS).

 

 

Thursday, 1 June

09:30 Registration & Welcome Coffee

Thursday morning: Opening Session

Convener: Jean-Paul Kneib (EPFL)

1 Welcome to the SKACH Spring Meeting

Welcome to the SKACH Spring Meeting in Versoix! This talk will provide some history of UniGe, and review the meeting schedule along with important housekeeping.

Speaker: Prof. Daniel Schaerer (University of Geneva)

2023_06_01_Consortium_Spring_MTG_Welcome.pdf

10:15 Q&A

2 SKACH Mid-Year Review

An overview of consortium activities since the winter meeting and a look forward to future activities in 2023.

Speaker: Carolyn Crichton

2023 Spring Meeting Mid-Year Update.pptx

10:35 Q&A

3 An Interactive Interferometer for Outreach

We present a 3D interactive visualization of a radio interferometer that was developed for outreach. This was deployed at the EPFL Open Days event on April 29-30, 2023.

Speaker: Shreyam Krishna (EPFL)

SKACH_Spring23_Radio_Imager.pdf

10:55 Q&A

Thursday morning: SKAO Regional Centers (SRCs)

Convener: André Csillaghy (University of Applied Sciences North Western Switzerland (FHNW))

4 Updates on Swiss SRC prototyping activities

This presentation will provide an update on the Coral Team's latest efforts in building SRCNet prototype services. The focus during the latest months has been on getting basic infrastructure up and running, and starting to deploy storage and metadata services that will be federated with other SRCNet site prototypes.

Speaker: Pablo Llopis Sanmillan (EPFL)

SKACH Spring Meeting 2023.pdf

11:15 Q&A

5 Computing Platforms and Infrastructure

The Computing Platforms and Infrastructure work package is responsible for enabling the science use cases for the SKACH consortium by providing an HPC platform at CSCS and contributing with technical expertise in HPC technologies, as well as software development.
In this presentation, we will show the status of the activities in the work package and provide insights about future activities.

Speaker: Victor Holanda (ETHZ/CSCS)

program_update_june_1st_2023_updated_version.pdf

11:35 Q&A

6 SKA Science Platform: Vision and Progress

Science analysis platforms will provide scientists with an interface to software and workflows, data processing, and analysis tools and hardware in a uniform way for massive science data products. It will involve one of the most significant existing science data lakes and a complex federated computational environment. This system should not be only powerful and efficient but, also, these extremely complex features should be presented to the community in a user-friendly way. In preparation for this, the SKA Regional Centre Steering Committee (SRCSC) has initiated several prototyping activities to investigate how currently existing tooling from other instruments and fields could be leveraged to provide the infrastructure within which the SRCs will operate.
Since the scale of science platform is humongous, the assessment of existing science platforms and discussion about the vision of the project is crucial. By assessing multiple existing science platforms, a future vision has been layed out. In this talk, we will talk about the activities of team Tangerine since Jan 2023. The feedbacks from the stakeholders and science expert has been obtained and implemented in the vision document. We also talk about creation of basic astronomical workflows in domain specific workflow managers like next flow etc..

Speakers: Rohit Sharma (FHNW, Windisch), Arpan Das (Scientific IT & Application Support, École polytechnique fédérale de Lausanne)

SKACH_spring_tagerine_team_update.pdf

11:55 Q&A

7 SRCSC WG6: Science User Engagement

The working group 6 of the SRC Steering Committee deals with Science User Engagement, providing feedback from the Science community to SKAO, in particular the Agile teams. A new JIRA ticketing system was set in place, and I will describe the process for feedback. Feedback can be made from Science Working Group representatives, WG6 members, and potentially external science users.

Speaker: Marc Audard (University of Geneva)

skach_mau_wg6_20230601.pdf

12:15 Q&A

12:20 SRC Discussion

Topic: SRC Synergies with SKACH Activities and Beyond

12:50 Lunch

Thursday afternoon: SKA Data Science Challenge

Convener: Michele Bianco (Swiss Federal Institute of Technology in Lausanne, EPFL)

8 SKA Data Challenge 3: Foreground removal update from SKACH team

Update from the participating SKACH team at the SDC3

Speaker: Dr Michele Bianco (Swiss Federal Institute of Technology in Lausanne, EPFL)

MBianco_SKACH_springmeeting.pdf

14:05 Q&A

9 Towards more realistic radio simulations

There is a large body of predictive research being done in preparation for the Square Kilometer Array. However, while the astrophysical models for these are usually well-informed, they fail to treat the radio interferometric signal capture and imaging process with the same vigour. The wide variety of modulations, both multiplicative (direction dependent, direction Independent and sub-categories thereof) and additive (For Eg: SEFD) are frequently ignored in favour of simply convolving simulation output with a Gaussian equivalent to the beam size of the synthesized beam of the telescope. We present some of the work we have undertaken for the SKA SDC3A, hoping to better account for these effects.

Speakers: Shreyam Krishna (EPFL), Chris Finlay (UNIGE)

Google Slides

SKACH_Summer_2023_Towards_Realistic_Mock_Observations.pdf

14:25 Q&A

Thursday afternoon: Data Science & Imaging

Convener: Emma Tolley (EPFL)

10 Conditional Denoising Diffusion Reconstruction of Radio Astronomical Images

In this work, we apply a novel approach for the reconstruction of radio astronomy images from uv-observations using guided diffusion. We use the Common Astronomy Software Applications package (CASA) to simulate data from Atacama Large Millimeter Array (ALMA). Our approach is based on denoising diffusion probabilistic models (DDPMs), which are effective at capturing the prior distribution given observations. We condition the model on the magnitude and phase of samples in the uv-space. By conditioning the model in this way, we are able to generate corresponding reconstructed images from noise realizations of the same size as reconstructed images and estimate both the localization and intensity of sources. The imaging process represents an information lossy operator, and thus the reconstruction is characterized by uncertainty. That is why, using a stochastic nature of the DDPMs, one can produce multiple realizations of reconstructed images for a given observation, which can be used for uncertainty estimation. To the best of our knowledge, this is the first time that DDPMs have been applied to radio UV data from ALMA generated using CASA. Our results demonstrate the potential of using stochastic generators for future analyses of radio astronomy data e.g., the Square Kilometer Array.

Speaker: Mariia Drozdova (UNIGE)

SKA - Sources localization and analysis-2.pdf

14:45 Q&A

11 Self-Supervised Learning for MeerKAT Images

Self-supervised techniques which leverage very large datasets have become standard in computer vision research, yet they are only now being applied to astrophysics. A major limitation is the characteristics of the data. In radio astronomy, flux values span a large dynamic range, and even the reduced size of the data in the image plane can be much larger than is typically used in computer vision problems.

This work addresses the simplified problem of radio continuum images, as observed or simulated to be observed by the SKA precursor MeerKAT. We experiment with various preprocessing steps, augmentations, and architectures to determine the optimal self-supervised learning approach for this type of data.

Our goal is to demonstrate that such techniques can produce meaningful embeddings which can serve as a starting point for many common data analysis tasks. Specifically, we examine object detection for continuum source detection, and similarity search.

Speaker: Erica Lastufka (University of Geneva)

SKACH_Spring_Lastufka.pdf

15:05 Q&A

12 Bluebild: a next generation radio interferometric imager

Bluebild is a next generation radio interferometric imager that reconstructs images directly on the celestial sphere. It does so by decomposing the captured visibilities into a set of eigenvalue-weighted eigen-visibilities via fPCA. We present updates on the scientific validation and development of this algorithm.

Speaker: Shreyam Krishna (EPFL)

SKACH_Spring23_Bluebild.pdf

15:25 Q&A

13 HVOX: Scalable Interferometric Synthesis and Analysis of Spherical Sky Images

Analysis and synthesis are key steps of the radio-interferometric imaging process, serving as a bridge between the visibility and sky domain. They can be expressed as partial Fourier transforms involving a large number of non-uniform frequencies and spherically-constrained spatial coordinates. Due to the data non-uniformity, these partial Fourier transforms are computationally expensive and represent a serious computational bottleneck in the image reconstruction process.

The W-gridding algorithm achieves log-linear complexity for both steps by applying a series of 2D non-uniform FFTs (NUFFT) to the data sliced along the so-called W frequency coordinate. A major drawback of this method however is its restriction to direction-cosine meshes, which are fundamentally ill-suited for large field of views. (Heavy pixel distortion as the radial distance increases.)

During the last SKACH meeting, we introduced HVOX, a novel algorithm for analysis/synthesis based on a 3D-NUFFT. Unlike W-gridding, the latter is compatible with arbitrary spherical meshes such as the popular HEALPix scheme for spherical data processing. Relying on the 3D-NUFFT allows one to optimally select the size of the inner FFTs, in particular the number of W-planes. This results in a better performing and user-friendly algorithm. To cope with the challenging scale of next-generation of radio telescopes, we proposed moreover a chunking strategy that partitions both the Fourier and image domains to balance the load amongst workers and distribute the 3D-NUFFT computational graph over multiple threads or processes with reduced memory requirements.

In this talk we present the latest developments related to HVOX, namely its new software interface and some novel benchmarking results demonstrating its scalability for both SKA and LOFAR. We also exemplify its use in the context of TART, a 24 element synthesis array radio-telescope observing the entire sky continuously and optimised for transient events detection (e.g., satellites, near-earth objects and high-energy cosmic rays).

Speakers: Sepand Kashani (AudioVisual Communications Laboratory (EPFL)), Dr Joan Rue Queralt (EPFL Center for Imaging), Dr Matthieu SIMEONI (EPFL Center for Imaging)

skach_spring_23.pdf

15:45 Q&A

15:50 Coffee Break & Discussion

Thursday afternoon: Science Enabling Tools

Convener: Simon Felix (FHNW)

14 Simulating radio observations with KARABO

KARABO is a versatile data simulator designed to handle the complex scientific requirements of the Square Kilometre Array (SKA) data. The simulator uses containerized packages and science-specific Python modules to generate mock data that allows for the study of scientific accuracy and the characterization of the instrument. KARABO's flexibility and parallelization make it a powerful tool for creating mock SKA data products, heliophysics requirements and simulating the EoR for SKA data challenge-3. This talk will discuss KARABO's implementation of simulating astronomical use cases from SKA and its precursor.

Speaker: Rohit Sharma (FHNW, Windisch)

ionsphere_screen.gif

summer_skach_meeting.odp

summer_skach_meeting.pdf

16:25 Q&A

15 Simulating HI Intensity Mapping for MeerKAT/SKA Mid

Neutral hydrogen (HI) intensity mapping makes it possible to
probe large volumes of the Universe. Since fluctuations of HI are a tracer for dark matter fluctuations, we can reconstruct the dark matter density field from HI intensity mapping. For HI intensity mapping systematics are a significant concern, therefore we need simulations to get a good understanding of their effects on the data. We are developing a simulation pipeline for interferometric intensity mapping with MeerKAT and SKA Mid. For that we use a sky catalog simulated by PINOCCHIO showing the HI intensity distribution. With Karabo we reconstruct the images we would get from an observation of line emission. This results in a three-dimensional HI intensity map which can be used to study
intermediate cosmological scales.

Speaker: Jennifer Studer

HI_IM_Karabo.pdf

16:45 Q&A

16 How Karabo supports Science users

We present Karabo user stories to demonstrate how we support science use cases. We talk about past examples where and how users requestes features, and what happened next.

Speaker: Simon Felix (FHNW)

2023-06-01 How Karabo Supports Science Users.pdf

17:05 Q&A

17 A GPU-accelerated and Python-wrapped update of the C2Ray radiative transfer code for Cosmic Reionization Simulations

The cosmic epoch of reionization (EoR) is studied using Radiative Transfer (RT) codes that simulate the radiative feedback from primordial galaxies and stars in the early Universe. This includes the simulation of the evolution of the neutral hydrogen fraction and thermal properties of the intergalactic medium (IGM) due to the computed ionizing and heating radiation. These simulations constitute the ground-basis experiment for the upcoming Square Kilometre Array (SKA) radio telescope. Ideally, they require huge volumes (∼ 1 Gpc size) and a large number of sources to reproduce the relevant cosmological scale. Therefore, RT and heating simulations require considerable computational power.

We propose to upgrade the C$^2$Ray code, widely used for EoR simulations. This code was written in Fortran90 in the late 90s (then updated over the years) and is CPU-parallelized using OpenMP and MPI. It uses short-characteristics raytracing to compute the propagation of UV radiation from primordial sources into the IGM and solves the differential equations that keep track of the neutral hydrogen fraction and thermal properties of the gas. The first part of the update consists in building a python interface to C$^2$Ray by wrapping its core subroutines as a fast compiled module. This could make its usage more flexible and allow some of its components to be used individually in other projects. The second part consists in upgrading the raytracing algorithm (which represents the dominant computational cost of the code) to make it GPU-parallelizable. The end goal is to achieve a significant reduction of the computation time required for EoR simulations.

Speaker: Patrick Hirling (EPFL)

SKACH_Spring_Hirling.pdf

17:25 Q&A

18 SKA HPC Co-design

The processing of a significant amount of data generated by the Square Kilometre Array (SKA) radio telescopes to produce scientific data products for answering complex radio astronomical questions requires high-performance supercomputers. However, the Science Data Processors (SDPs), which are one of the primary processing facilities of SKA, have limited power resources. Thus, it is crucial to evaluate different hardware solutions and software optimizations to minimize procurement risks for the SDP.

To this end, the SKA CO-design and OPtimisation (SCOOP) team is working on the SW/HW co-design for the SDP. The presentation will discuss the main activities performed by the SCOOP team, including the development of the SDP benchmark suite for hardware evaluation, initial study and profiling of the calibration pipelines, and benchmarking and optimization of radio-astronomical algorithms developed by the SKA Software Teams.

The presentation will conclude by outlining SCOOP's work plan for the next phase, which involves enhancing the SDP benchmark suite and benchmarking new and promising hardware for the SDP.

Speaker: Dr Stefano Corda (EPFL)

01062023-codesign-skach-spring.pptx

17:45 Q&A

Closing remarks

18:00 Apero

Friday, 2 June

Friday Morning: Data Science & Simulations

Convener: Darren Reed (University of Zurich)

19 Sky Simulations update

I will give an overview of the Sky Simulations work package, including a summary of recent activities, future plans and goals.

Speaker: Darren Reed (University of Zurich)

reed_skysims_june2023.pdf

09:15 Q&A

20 First results of the Phoebos Simulation

I'll discuss the first results from the Phoebos Simulation

Speaker: Floor van Donkelaar

skachspring.pptx

09:35 Q&A

21 Field level model for HI: mocks and applications

I will summarise results on modelling HI from different hydro simulations using analytical field level model. I will then give an update on the method and the code 'Hi-Fi mocks' to generate arbitrary large 3D HI mocks from smaller hydro-simulations. Finally I will show how robustly one can recover the input cosmology from these mocks using power spectrum and use these results to forecast the cosmological constraints from upcoming 21cm surveys.

Speaker: Andrej Obuljen

Obuljen_forward_model_Versoix.pdf

09:55 Q&A

22 Deep learning image-to-image translation for cosmological simulations and SKA mock observations

In the study of the Universe, particularly galactic dynamics,
hydrodynamical simulations have played a critical role. Finely-tuned on a wide range of scales, recent large-scale simulation suites have produced mock observations that are nearly indistinguishable from actual observations.

In comparison to traditional machine learning techniques, the
complexity of these models poses a substantial challenge. However, artificial neural networks have the ability to learn by transforming data iteratively into a more compressed and abstract representation. Once optimized, these networks can generalize to unseen data, which makes them more applicable in a "real-world" setting.

In this talk, we would like to provide a status update on the current effort at ZHAW to explore the mapping from cosmological simulations to mock images from simulated SKA visibilities (and vice versa), using deep learning techniques. We will present a comparison of the performance of two image-to-image translation conditional adversarial network models, pix2pix and CycleGAN, trained on a dataset of galaxies from the IllustrisTNG simulation suite.

Speaker: Dr Philipp Denzel (Centre for Artificial Intelligence, ZHAW)

Denzel_SKACH_spring_23.pdf

Image-to-image translation between SPH simulations and SKA mocks

10:15 Q&A

23 Trajectory Based RFI Subtraction and Calibration

We present 𝚝𝚊𝚋𝚊𝚜𝚌𝚊𝚕, (𝐓r𝐀jectory 𝐁𝐀sed RFI 𝐒ubtraction and 𝐂𝐀𝐋ibration), a new algorithm that jointly models the RFI signal, trajectory, antenna gains, and residual visibilities to recover RFI-free interferometry data. We show that 𝚝𝚊𝚋𝚊𝚜𝚌𝚊𝚕 improves image quality beyond what is achieved with the standard flagging approach as well as existing RFI subtraction techniques. 𝚝𝚊𝚋𝚊𝚜𝚌𝚊𝚕 has been further improved to overcome scalability and generalisation issues that had previously held it back.

Speaker: Chris Finlay (UNIGE)

Google Slides

TABASCAL_SKACH_Spring_Meeting_2023.pdf

10:35 Q&A

10:40 Coffee Break & Discussion

Friday Morning: SKACH Science

Convener: Prof. Daniel Schaerer (University of Geneva)

24 Radio Continuum Observations of analogues of the Sources of Cosmic Reionization

The nature of sources that led to the last phase transition of matter in the Universe, namely the cosmic reionization, remains an open question in astrophysics. It is widely believed that low-mass star-forming galaxies (emission lines galaxies, Lyman alpha emitters and related objects) are the major contributors to reionization. Thus they have been intensely studied at multiple wavelengths using state-of-the-art facilities including HST, 10m-class telescopes and others. Recently the first large sample of Lyman-Continuum (LyC) emitting galaxies has been found at low redshift (z~0.3) with HST and the SDSS, revealing objects with strong ionising hoton production and high LyC photon escape fractions (fesc), which are cosmologically relevant, and with properties making them the best analogues of the sources of cosmic reionization.

To better understand the nature, the physical processes at play, and the multi-wavelength SEDs of these interesting galaxies we have undertaken the first radio continuum (RC) observations of such objects. We will present the first results from this study using the Very Large Array (VLA) multiple radio bands (1.5-8 GHz). We have also observed a subset
of the sample using the GMRT at low frequencies (0.3-1.2 GHz).

We have detected the RC in a sample of Low-z LyC emitting galaxies. Their radio-spectral energy distribution (SED) at GHz shows a wide variety along with evidence for spectral steepening and free-free absorption. We find a tentative correlation between the radio spectral index and fesc. Furthermore, our galaxies show a deviation from the
standard radio-SFR relation. These multi-frequency radio observations provide unique insights into the physical properties of the interstellar medium and feedback
processes in these galaxies which are in turn related to LyC escape.

Our observations will provide guidance for future multi-wavelength studies of high-redshift galaxies and searches for LyC emitters with upcoming highly sensitive facilities like the SKA.

Speaker: Dr Omkar Bait (University of Geneva)

Omkar_Bait_SKACH_Spring_meeting_23.pdf

11:15 Q&A

25 The Hydrogen Intensity and Real-time Analysis eXperiment (HIRAX) - Overview and Status

The Hydrogen Intensity and Real-time Analysis eXperiment (HIRAX) is a radio interferometer array under development and to be deployed at the South African Radio Astronomy Observatory (SARAO) Square Kilometer Array (SKA) site in South Africa. HIRAX is entering in the deployment phase of the initial 256-element array of 6m dishes. I will present an overview of the HIRAX instrument and planned survey as well as updates on the status of the X-Engine correlator and Science Data Processing (SDP) systems being commissioned in Switzerland.

Speaker: Devin Crichton (ETH Zurich)

Devin Crichton - SKACH Summer Meeting .pdf

11:35 Q&A

26 21cm foreground removal with machine learning

Observation of the 21cm signal through intensity mapping is a promising probe of large-scale structures in the universe. The central challenge of 21 cm intensity mapping is that the method is highly sensitive to the bright astrophysical foregrounds. Traditional component separation requires exquisite spectral purity to subtract the foreground components with smooth spectral responses from the HI signal with fluctuating spectral features. However, such methods are sensitive to systematics that distort the spectral smoothness, resulting in foreground residuals in the cleaned data. In this work, we explore the possibility of using machine learning to subtract both foreground and systematics from data. We test our machine learning network on a number of sky simulation models and instrumental effects. We compare our foreground subtraction results with those obtained using traditional approaches like the principle component analysis (PCA). We found that our network outperforms PCA in terms of signal loss and the results are robust against different sky models and instrumental responses.

Speaker: Tianyue Chen (EPFL)

ML_FG.pdf

11:55 Q&A

27 Headline science in band 6

I will outline the key science drivers of SKA1-MID observations in the frequency range 15-25 GHz (and in some cases up to 50 GHz). Additionally I will summarize the "observatory landscape" (e.g., ngVLA, ALMA band 1, synergies with other southern hemisphere facilities) such a high-frequency capability would operate in. Finally, this talk would also aim to garner momentum for a document describing the interest of the Swiss scientific community in a MID B6 capability, building on the white paper already written by the SKA community at large in 2019/20 ("SKA1 Beyond 15GHz:
The Science case for Band 6"; SKA memo 20-01).

Speaker: Dr Mark Sargent (International Space Science Institute (ISSI))

SKAMID-B6_MTSargent_020623.pdf

12:15 Q&A

12:30 AstroSignals Signals Sinergy Discussion

Topic: Planned, ongoing and foreseen synergies

13:00 Lunch & Closing

Board Meeting: Board Members Only