Periodic Reporting for period 1 - TITAN (Frugal Artificial Intelligence and Application in Astrophysics)
Periodo di rendicontazione: 2023-01-01 al 2024-03-31
In our awe-inspiring quest to comprehend the universe, to explore the very genesis of the cosmos, to shed light on the essence and distribution of dark matter, to seek out signs of life on distant worlds, and perchance to discover that we are not alone, astrophysicists increasingly lean on mathematics, statistics, and data-driven machine learning. As a result, Astroinformatics has emerged as a swiftly expanding discipline that amalgamates contemporary computational techniques, sophisticated data analysis algorithms, streamlined software implementations, and corresponding hardware design. This integration enables the observation of new phenomena and the formulation of predictions in astronomy, fostering data-driven breakthroughs in astronomical research.
The TITAN project has selected Dr. Jean Luc Starck, an esteemed researcher, as the ERA Chair holder of Astroinformatics. Dr. Starck will spearhead a research initiative in computational astrophysics and data science, aimed at synergizing and enhancing two major research entities at FORTH: Computer Science and Astrophysics. This initiative will establish a robust link between regional research excellence and international endeavors. Our goal is to establish an interdisciplinary, international, and culturally diverse center of excellence focused on advancing computational science, machine learning, and statistics for research and education in data-oriented Astronomy. TITAN aims to unleash the potential of FORTH, the Cretan region, and Greece through various measures including capacity building (expansion of research programs, provision of advanced training opportunities), institutional development and structural reforms, and fostering intersectoral and international partnerships in the realm of Computational Astrophysics.
The TITAN project has selected Dr. Jean Luc Starck, an esteemed researcher, as the ERA Chair holder of Astroinformatics. Dr. Starck will spearhead a research initiative in computational astrophysics and data science, aimed at synergizing and enhancing two major research entities at FORTH: Computer Science and Astrophysics. This initiative will establish a robust link between regional research excellence and international endeavors. Our goal is to establish an interdisciplinary, international, and culturally diverse center of excellence focused on advancing computational science, machine learning, and statistics for research and education in data-oriented Astronomy. TITAN aims to unleash the potential of FORTH, the Cretan region, and Greece through various measures including capacity building (expansion of research programs, provision of advanced training opportunities), institutional development and structural reforms, and fostering intersectoral and international partnerships in the realm of Computational Astrophysics.
The scientific and technical endeavors of TITAN have concentrated on pioneering a radically innovative approach to inverse problems in astrophysics. This approach is characterized by its efficiency, reliability, scalability with measurement dimensions and volumes, and its ability to enhance image resolution and quality. To address the computational bottleneck and tackle scalability concerns, the ERA Chair has introduced a large-scale data processing methodology. Deep Learning (DL) plays a pivotal role in enhancing and expediting algorithms, thereby profoundly impacting targeted applications. The ERA Chair has been instrumental in promoting advancement across six key challenges.
• Creation of a resilient DL framework: Our team devises innovative DL strategies to regulate inverse problems, employing a physics-centric methodology that exhibits strong generalization to unfamiliar datasets during training. This framework offers quantification of uncertainties while maintaining efficiency, thereby minimizing the environmental impact of computations and data storage.
• Expanding the application of DL to higher-dimensional data: We harness the full potential of DL to integrate observations from diverse sensing modalities, even in high-dimensional formats.
• Improving radio-interferometry image reconstruction: We devise an effective approach to capitalize on the widefield imaging capabilities, high angular resolution, and unparalleled instantaneous sensitivity of SKA (Square Kilometre Array) to detect and characterize radio emissions from sources.
• Creating resilient weak lensing techniques for optical and radio observations: We design novel tools for radio weak lensing image reconstruction, as well as galaxy shape and mass mapping methods tailored for the Euclid and SKA missions.
• Promote frugality in Astrophysics: We address the challenge of resource-intensive GPU N-body simulations in likelihood-free cosmological parameter inference.
• Advancement of state-of-the-art technologies: Our algorithmic framework facilitates the creation of groundbreaking technologies such as AI-driven sensing platforms, distributed camera networks for tracking space assets, and innovative space mission designs for both astrophysical research and Earth Observation.
• Creation of a resilient DL framework: Our team devises innovative DL strategies to regulate inverse problems, employing a physics-centric methodology that exhibits strong generalization to unfamiliar datasets during training. This framework offers quantification of uncertainties while maintaining efficiency, thereby minimizing the environmental impact of computations and data storage.
• Expanding the application of DL to higher-dimensional data: We harness the full potential of DL to integrate observations from diverse sensing modalities, even in high-dimensional formats.
• Improving radio-interferometry image reconstruction: We devise an effective approach to capitalize on the widefield imaging capabilities, high angular resolution, and unparalleled instantaneous sensitivity of SKA (Square Kilometre Array) to detect and characterize radio emissions from sources.
• Creating resilient weak lensing techniques for optical and radio observations: We design novel tools for radio weak lensing image reconstruction, as well as galaxy shape and mass mapping methods tailored for the Euclid and SKA missions.
• Promote frugality in Astrophysics: We address the challenge of resource-intensive GPU N-body simulations in likelihood-free cosmological parameter inference.
• Advancement of state-of-the-art technologies: Our algorithmic framework facilitates the creation of groundbreaking technologies such as AI-driven sensing platforms, distributed camera networks for tracking space assets, and innovative space mission designs for both astrophysical research and Earth Observation.
The discoveries made by TITAN enhance the attractiveness of FORTH and the Crete region, drawing in young talents who typically opt for institutions in the US or prominent establishments in Northern Europe. Acquiring advanced knowledge through TITAN not only enhances employment opportunities in the Southeastern Mediterranean, Europe, and globally but also satisfies the considerable demand for such expertise in both academic and industrial sectors.
In alignment with the Rebrain Greece initiative recently launched by the Greek Government, TITAN serves as a catalyst in reversing brain drain by fostering a competitive environment that encourages the return of highly skilled Greek scientists from overseas.
TITAN actively contributes to national and regional training and educational initiatives by implementing measures such as introducing new University courses, organizing summer schools, hosting seminars, and conducting workshops.
TITAN leads an innovation initiative focused on involving industrial and business stakeholders through activities, such as scientific seminars, training programs, industry weeks, and collaborative academia-industry research projects. Regional industrial stakeholders and governmental departments are encouraged to participate in outreach events and industry weeks, promoting awareness of the applications of computational intelligence and data science beyond Astrophysics.
In alignment with the Rebrain Greece initiative recently launched by the Greek Government, TITAN serves as a catalyst in reversing brain drain by fostering a competitive environment that encourages the return of highly skilled Greek scientists from overseas.
TITAN actively contributes to national and regional training and educational initiatives by implementing measures such as introducing new University courses, organizing summer schools, hosting seminars, and conducting workshops.
TITAN leads an innovation initiative focused on involving industrial and business stakeholders through activities, such as scientific seminars, training programs, industry weeks, and collaborative academia-industry research projects. Regional industrial stakeholders and governmental departments are encouraged to participate in outreach events and industry weeks, promoting awareness of the applications of computational intelligence and data science beyond Astrophysics.