Quantum Computational Intelligence

Abstract: The world of computing is going to shift towards new paradigms able to provide better performance in solving hard problems than classical computation. In this scenario, quantum computing is assuming a key role thanks to the recent technological enhancements achieved by several big companies in developing computational devices based on fundamental principles of quantum mechanics: superposition, entanglement and interference. These computers will be able to yield performance never seen before in several application domains, and the area of artificial intelligence may be the one most affected by this revolution. Indeed, on the one hand, the intrinsic parallelism provided by quantum computers could support the design of efficient algorithms for artificial intelligence such as, for example, the training algorithms of machine learning models, and bio-inspired optimization algorithms; on the other hand, artificial intelligence techniques could be used to reduce the effect of quantum decoherence in quantum computing, and make this type of computation more reliable. This talk aims at introducing the auditors with this new research area and pave the way towards the design of innovative computing infrastructure where both quantum computing and artificial intelligence take a key role in overcoming the performance of conventional approaches.

Short Bio: Giovanni Acampora: Professor of Quantum Machine learning at the University Federico II and coordinator of the Ph.D. program in Computational Intelligence. He introduced the Fuzzy Markup Language (FML)which became the first IEEE standard  (IEEE-1855) in Fuzzy logic and gained him the IEEE-SA Emerging Technology Award. He is Editor-in-Chief of the Journal Quantum Machine Intelligence and in 2019,  he received the Canada-Italy Innovation Award.

A BRAIN study to tackle image analysis with artificial intelligence in the ALMA2030 era

Abstract: An ESO internal ALMA development study, BRAIN is addressing the ill-posed inverse problem of image analysis employing astrostatistics and astroinformatics. These emerging fields of research offer interdisciplinary approaches at the intersection of observational astronomy, statistics, algorithm development, and data science. In this study, we provide evidence of the benefits in employing these approaches to ALMA image analysis for operational and scientific purposes. We show the potentials of two techniques (RESOLVE and DeepFocus), applied to ALMA calibrated science visibilities. Significant advantages are provided with the potential to improve the quality and completeness of the data products and overall processing time. Both approaches evidence the logical pathway to address the incoming revolution in data analysis dictated by ALMA2030. Moreover, we bring to the community additional products through a new package (ALMASim) to promote advancements in these fields, providing a refined ALMA simulator usable by a large community for training and/or testing new algorithms.

Short Bio: As a Scientist at the ALMA Regional Centre at the European Southern Observatory (ESO HQ, Garching), she serves as the European Data Reduction Manager and leads the principal investigator role for the ESO internal ALMA development study, BRAIN, aimed at enhancing CASA imaging procedures. Her responsibilities encompass quality assurance, pipeline testing, and subsystem development to enhance the operational workflow utilized by several data reducers within the ALMA partnership. Fabrizia’s interests are centered around astrostatistics and astroinformatics, multiwavelength research, along with creating astrometric catalogues. She actively contributes to popularizing Bayesian techniques (see ‘Bayes Forum’ in Garching research campus) which offer innovative solutions to tackle the challenges posed by big data in astronomy. Through the utilization of advanced algorithms that optimize decisions based on well-informed assumptions, she addresses these challenges with refined machine learning techniques. Her focus is to introduce these innovative techniques in the current ALMA data reduction workflow.

Anomaly detection: overview and application to ZTF

Abstract: Optical astronomy has become increasingly data rich with billions of sources and hundreds of observations per source. Many methods are being planned to use data-driven ways to classify objects, be they in the Solar System, or variable stars in our Galaxy, or extra-galactic objects. While current surveys and experience over a few decades has prepared us for well understood classes, data covering wider parameter spaces are likely to harbor classes and phenomena not encountered before. There would be in-class extreme outliers, and totally new phenomena. Identifying all kinds of anomalies is crucial to advance our understanding of the cosmos. The most exciting anomalies would be those that reveal our biases and selection effects, and lead us to entire populations unknown so far. We are also prepared for finding a lot of artifacts in the process. A bit like bugs, these could reveal shortcomings in the processing pipelines, and an opportunity for improvement. We will provide a broad summary of anomaly detection and present our work on finding anomalies in ZTF data starting from light curve features and methods like HDBSCAN and isolation forest to look for outliers. We will finish by showing how this connects with the bigger picture, and can be generalized for other surveys.

Short Bio: Ashish Mahabal is an astronomer (Division of Physics, Mathematics, and Astronomy) and Lead Computational and Data Scientist (Center for Data Driven Discovery) at the California Institute of Technology. His interests include Large Sky Surveys, Classification, Deep Learning, and Methodology Transfer to other complex-data fields like medicine. He leads the ML for the Zwicky Transient Facility, a new large survey covering the entire Northern Sky every few nights. He also works with the Data Science group at the Jet Propulsion Laboratory and is part of the Early Detection Research Network (EDRN) for cancer, and MCL.

From Photometric Redshifts to Improved Weather Forecasts: an interdisciplinary view on machine learning in astronomy

Abstract: The amount, size, and complexity of astronomical data-sets is growing rapidly in the last decades. Now, with new technologies and dedicated survey telescopes, the databases are even growing faster. Besides dealing with poly-structed and complex data, sparse data has become a field of growing scientific interest. By applying technologies from the fields of computer sciences, mathematics, and statistics, astronomical data can be accessed and analyzed more efficiently.
A specific field of research in Astroinformatics is the estimation of the redshift of extra-galactic sources, a measure of their distance, by just using sparse photometric observations. Observing the full spectroscopic information that would be necessary to directly measure the redshift, would be too time-consuming. Therefore, building accurate statistical models is a mandatory step, especially when it comes to reflecting the uncertainty of the estimates. Statistics and especially weather forecasting has introduced and utilized proper scoring rules and especially the continuous ranked probability score to characterize the calibration as well as the sharpness of predicted probability density functions.
This talk presents what we achieved when using proper scoring rules to train deep neural networks and to evaluate the model estimates. We present how this work led from well calibrated redshift estimates to an improvement in statistical post-processing of weather forecast simulations. The presented work is an example of interdisciplinarity in data-science and how methods can bridge between different fields of application.

Short Bio: Kai Polsterer is the Astroinformatics Group leader and Deputy Scientific Director at the Heidelberg Institute for Theoretical Studies operated by the Klaus Tshira Foundation. He is one of the European leaders in the field of astroinformatics. His contributions span a wide range of fields from astrophysics, to weather forecast, to medicine.  

Contrastive learning applied to astrophysics

Abstract: Reliable tools to extract patterns from high-dimensionality spaces are becoming more necessary as astronomical datasets increase both in volume and complexity. Contrastive Learning is a self-supervised machine learning algorithm that extracts informative measurements from multi-dimensional datasets, which has become increasingly popular in the computer vision and Machine Learning communities in recent years. To do so, it maximizes the agreement between the information extracted from augmented versions of the same input data, making the final representation invariant to the applied transformations. Contrastive Learning is particularly useful in astronomy for removing known instrumental effects and for performing supervised classifications and regressions with a limited amount of available labels. I will summarize the main concepts behind contrastive learning and review the first promising applications to astronomy. I also include some practical recommendations on which applications are particularly attractive for contrastive learning and present some examples on integral field spectroscopic data.

Short Bio: Born in Argentina, she is finishing her PhD on Galaxy Evolution and Deep Learning at the Institute for Astrophysics of the Canaries (IAC, Spain). Her work combines integral field spectroscopic (IFS) data and cosmological simulations with Deep Learning techniques to study the galaxy assembly history.

Connected Morphological Operators in Astronomy: Tools for Object Detection and Pattern Analysis on Vast Data.

Abstract: Connected morphological filters have seen rapid development theory, algorithms, and applications in many fields of computer vision. More recently, they have found use in astronomical applications, most notably in object detection and pattern analysis. Because these operators work on connected regions in the image in a data-driven way, rather than applying some arbitrary kernel to all pixels equally, parallel algorithms were initially not available. In the past decade, this situation has been amended, and parallel and distributed algorithms capable of handling giga- and tera-scale data sets are available. In this talk I will give an overview of existing tools and algorithms, and new developments in multi-band object detection, dealing with huge data cubes from e.g. LOFAR, exploratory data analysis, and combining these methods with machine-learning tools.

Short Bio: Michael Wilkinson obtained an MSc in astronomy from the Kapteyn Astronomical Institute, University of Groningen in 1993, after which he worked on image analysis of intestinal bacteria at the Department of  Medical Microbiology, University of Groningen, obtaining a PhD at the Institute of Mathematics and Computing Science, also in Groningen, in 1995. He was appointed as researcher at the Centre for High Performance Computing in Groningen working on simulating the intestinal microbial ecosystem on parallel computers. After this he worked as a researcher at the Johann Bernoulli Institute for Mathematics and Computer Science on image analysis of diatoms. He is currently associate professor at the Bernoulli Institute for Mathematics, Computer Science and Artificial Intelligence, working on morphological image analysis and especially connected morphology and models of perceptual grouping. An important research focus is on handling giga- and tera-scale images in remote sensing, astronomy and other digital imaging modalities.

Unleashing the Power of Transformers for the Analysis of Cosmic Streams

Transformers are deep learning architectures that have shown to reach state-of-the-art performances across various domains. They originally gained attention thanks to their performance in natural language processing tasks. More recently, they have been successfully applied to tasks involving images, tabular data, and time series, among others. In this talk we will review the recent advancements in transformers when applied to the characterization of astronomical light-curves, from task-specific models to foundation models. Transformers have become the new state-of-the-art and will play a key role in analysing cosmic streams in real time from current and next-generation time-domain instruments such as the Vera C. Rubin Observatory and its Legacy Survey of Space and Time (LSST).

Short Bio: Associate Professor at the Department of Computer Science, University of Concepción. His research interests include machine learning, computer vision, data-science, astroinformatics, and bioinformatics. His work is focused on developing new algorithms for massive observational data from a wide variety of astronomical instruments such as ALMA, SDSS, HST, and the LSST among others. he is a founding member of the Astroinformatics Laboratory at the Center for Mathematical Modeling, and is currently a member of the Millenium Institute of Astrophysics.

Solving Large-scale Data Challenges with ESA Datalabs

Current and upcoming space science missions will produce petascale data in the coming years. This requires a rethinking of data distribution and processing practices. For example, the Euclid mission will be sending more than 100GB of compressed data to Earth every day. Analysis and processing of data on this scale requires specialized infrastructure and toolchains. Further, providing users with this data locally is not practical due to bandwidth and storage constraints. Thus, a paradigm shift of bringing users’ code to the data and providing a computational infrastructure and toolchain around the data is required. The ESA Datalabs platforms is specifically focused on fulfilling this need. It provides a centralized platform with access to data from various missions including the James Webb Space Telescope, Gaia, and others. Pre-setup environments with the necessary toolchains and standard software tools such as JupyterLab are provided and enable data access with minimal overhead. And, with the built-in Science Application Store (SCIAPPS), a streamlined environment is given that allows rapid deployment of desired processing or science exploitation pipelines.  In this manner, ESA Datalabs provides an accessible and potent framework for high-performance computing and machine learning applications. While users may upload data, there is no need to for downloading data, thus mitigating the bandwidth burden. As the computational load is handled within the computational infrastructure of ESA Datalabs, high scalability is achieved, and resources can be requisitioned as needed. Finally, the platform-centric approach facilitates direct collaboration on code and data. Currently, the platform is already available to several hundred users, regularly showcased in dedicated workshops and interested users may request access online.

Short Bio: Tba

Training on the interface of astronomy and computer science

Abstract:Modern developments in fields like astronomy imply the need for advanced computational approaches to handle large data sets. However, most recent and current PhDs in astronomy have received little or no formal training in important areas including computer science, computational methods, software and algorithm design, or project management.
We will report on a number of scientific advances on the interface of extragalactic astronomy and computer science that have resulted from our EU-funded Initital Training Network SUNDIAL, in which 15 PhD candidates were trained. We will then discuss how these results will be built upon and expanded in a newly approved Marie-Sklodowska Curie Doctoral Network called EDUCADO (Exploring the Deep Universe by Computational Analysis of Data from Observations). We will train 10 Doctoral Candidates in the development of a variety of high-quality methods needed to address the formation of the faintest observed structures in the nearby Universe, including novel object detection algorithms and object recognition and parameter distribution analyses of complex datasets. We aim to detect unprecedented numbers of the faintest observable galaxies from new large-area surveys. We will study the morphology, populations, and distribution of large samples of various classes of dwarf galaxies, compare dwarf galaxy populations and properties across different environments, and confront the results with cosmological models of galaxy formation and evolution. Finally, we will perform detailed simulations and observations of the Milky Way and the Local Group to compare with dwarf galaxies in other environments. We aim for our interdisciplinary training tools, methods and materials to become publicly available.

Short Bio: tba

Physics Informed Neural Networks (PINNS); Solving Differential Equations Using Neural Networks.

Abstract: In this talk, I will review how we can use neural networks to solve differential equations for initial conditions and boundary value problems. There are many exciting directions in this line of work, but two main criticisms are computational complexity compared to traditional methods and the lack of error bounds. I will present transfer learning methods that can be used to speed up computation and new work on estimating the error bounds of the neural network approach in solving (partial) differential equations.

Short Bio: tba

Navigating Global Policy and Diplomacy through Astroinformatics: Insights from Science20

Abstract: In the ever-evolving nexus of global policy and technical innovation, the burgeoning field of astroinformatics emerges as a beacon of collaborative potential. As we stand at the intersection of scientific discovery, policy formulation, and international diplomacy, the Science20 (S20) platform serves as a bridge uniting technical expertise from around the world with policy discours. Astroinformatics, at its core, encapsulates the convergence of astronomy, data science, and cutting-edge technologies. Within this context, the recent Science20 engagement titled “Astroinformatics for Sustainable Development” stood as a pivotal milestone. This symposium, held virtually on July 6th and 7th, 2023, delved into the global panorama of astroinformatics and its profound impact on policy and diplomacy. In this talk, I will present the policy discourse that emerged from the Science20 platform highlighting the various views and translatable policy and science diplomacy frameworks.

Short Bio: Pranav Sharma is an astronomer and science historian known for his work on the history of the Indian Space Program. He has curated Space Museum at the B. M. Birla Science Centre (Hyderabad, India) and led several exhibitions on Indian Space History in collaboration with ISRO, CNES, ESA, and European Union Institute. He was the in-charge of the history of the Indo-French scientific partnership project supported by the Embassy of France in India. He is the Co-Curator of the History of Data-Driven Astronomy Project. He also serves as the Policy and Diplomacy Advisor to United Nations International Computation Centre, Advisor to the France India Foundation, Scientific Advisor to Arc Ventures, and Member Secretary (Policy, Transdisciplinary Disruptive Science, and Communications) for G20-Science20.  He has co-authored the book Essential Astrophysics: Interstellar Medium to Stellar Remnants, CRC Press, 2019.

Rapid Inference for Extragalactic Transients  in the Era of LSST

Soon, the Legacy Survey of Space and Time (LSST) will commence and drive the discovery rate of extragalactic transients to millions per year. Much work has been done to identify these transients, to classify transients in realtime, and to aid researchers in deciding how to optimize observational resources to characterize these events. Here, I will discuss recent developments in inference (i.e., extracting astrophysical information) for such events. I will focus on the use of emulators to approximate complex physical simulations, and simulation-based inference techniques to accelerate Bayesian inference.

Short Bio: Ashley Villar is Assistant Professor of Astronomy at Harvard University. Previously, she was the inaugural Mercedes Richards Career Development Assistant Professor at the Pennsylvania State University. Her main interests are in using data-driven methods and machine learning to study the eruptions, mergers and explosions of stars. She is especially interested in utilizing multiband light curves to understand the underlying physics of optical transients. 

From JWST to Euclid: new algorithms  for the study of the first galaxies

Abstract: Just-launched space missions like JWST and Euclid have just been launched and are already revolutionising our understanding of many physical phenomena in the local and the distant Universe.

In this talk I will concentrate on the field where the synergy between these two missions is the strongest: the search and study of the evolution of galaxies, starting from the first that appeared shortly after the Big Bang.

After a short introduction of the current status of this research field, I will discuss the challenges in the interpretation and analysis of the data that are obtained with these new facilities.

The large amount of data demand automated procedures for their analysis, and the subtle systematics that may easily affect or bias the results require the development of AI or other advanced techniques to be avoided.

I will briefly describe a few test cases where the adoption of advanced algorithms may improve the reliability and accuracy of the analysis.

Short Bio: Research director at INAF, president of theLBT corporation and a world authority in the field of high redshift galaxy evolution. He had leading responsabilities in most current and future survey projects such as K20, GOODS, CANDELS, VUDS VANDELS, Frontier Fields, EUCLID, etc. Since a few years his group has been proposing innovative AI based approaches to the study of the high redshift universe. He currently leads the 70 Meuro STILES project which will drive italian astronomy in the next generation of observing and computing facilities.

Developments in Fast Machine Learning for Science 

For over two decades, we’ve heard about the challenges that the forever imminent data tsunami is going to bring to astronomy and yet we still spend most of our computing lives firmly in the land of CPUs with occasional forays into the GPU realm. There are, however, now genuine needs across the entire science spectrum for high throughput low latency inferencing driven by real-time responses that our current systems cannot deliver. Novel options in both hardware components and algorithmic methodologies are required to deliver production-ready solutions that can cope. In this talk, I will review current developments in the field of fast machine learning and, in particular, those that are relevant to multimessenger astronomy which is at the forefront of this work in our domain.

Short Bio: He is Research Professor of Astronomy at the California Institute of Technology and the Project Scientist for the Zwicky Transient Facility (ZTF), the first of a next generation of time-domain sky surveys producing hundreds of thousands of public transient alerts per night. Previously he has worked on the Catalina Real-time Transient Survey (CRTS), a still unmatched data set in terms of temporal baseline coverage; the NOAO DataLab; the Virtual Observatory; and the Palomar-Quest Digital Sky Survey. He is member of the NSF-funded HDR Institute for Accelerated AI Algorithms for Data-Driven Discovery (A3D3) which aims to provide real-time AI at scale in high energy physics, multimessenger astronomy, and neuroscience. In this field, his particular interest is in low-latency inferencing from astronomical alert streams with commodity hardware accelerators and using reinforcement learning to optimize followup strategies. His main research interests are the application of machine learning and advanced statistical methodologies to astrophysical problems, particularly the variability of quasars and other stochastic time series.

Enabling New Discoveries with Machine Learning

Abstract: The next generation of telescopes such as the SKA and the Vera C. Rubin Observatory will produce enormous data sets, far too large for traditional analysis techniques. Machine learning has proven invaluable in handling massive data volumes and automating many tasks traditionally done by human scientists. In this talk, I will explore the use of machine learning for automating the discovery and follow-up of interesting astronomical phenomena. I will share an exciting recent MeerKAT discovery made with machine learning and discuss how the human-machine interface will play a critical role in maximising scientific discovery with automated tools.

Short Bio: Senior Lecturer with a joint position between the University of the Western Cape and the South African Radio Astronomy Observatory (formerly SKA South Africa). Her focus is on cosmology and trying to get the best out of combining optical and radio telescopes like the Vera C. Rubin Observatory, in Chile, as well as the Square Kilometre Array and its precursor, MeerKAT, in South Africa. She works on developing new statistical techniques and machine learning techniques with special focus on anomaly detection for scientific discovery. She is also the founder and director of an international mentoring programme for women and gender minorities in physics called the Supernova Foundation.

Quantum computing for natural sciences and machine learning applications

Abstract:Over the last few decades, quantum information processing has emerged as a gateway towards new, powerful approaches to scientific computing. Quantum technologies are nowadays experiencing a rapid development and could lead to effective solutions in different domains including physics, chemistry, and life sciences, as well as optimization, artificial intelligence, and finance. To achieve this goal, noise-resilient quantum algorithms together with error mitigation schemes have been proposed and implemented in hybrid workflows with the aim of improving the synergies between quantum and classical computational platforms. In this talk, I will review the state-of-the-art and recent progress in the field, both in terms of hardware and software, and present a broad spectrum of potential applications, with a focus on natural sciences and machine learning.

Short Bio: tba

High-Dimensional Data Visualization and AI-Enhanced Exploratory Data Analysis in a Commercial Software Solution

The talk will cover the innovative techniques for exploring and analyzing high-dimensional data within the context of a commercial visualization software. I will start with the challenges of rendering network graphs, scatter plots, and a multitude of other data representations on both desktop and VR platforms, while aiming to offer ergonomic user interaction and visualization.

I will continue by introducing the AI-driven analytical tools that the proposed solution provides for identifying and analyzing complex network structures and data patterns. The presented methods are versatile and capable of automatically extracting insights from diverse multi-dimensional datasets—ranging from categorical and numerical data to natural language data—across a wide spectrum of domains; the versatility comes with inevitable trade-offs that I hope will ignite discussions on how to effectively mitigate these challenges, especially at scale.

Short Bio: He is currently serving as a Senior Machine Learning Contractor at Virtualitics, Inc., a role he has held since 2021 while he earned his M.S. (summa cum laude) degree in Computer Science from the University of Salerno. He shortly held a research associate position at the University of Portsmouth in 2019 and is currently pursuing his Ph.D. in the field of digital twins for territory management and disaster response at the University of L’Aquila.