Avsnitt
-
In this episode, we interview Logan Kilpatrick. Logan currently splits his time between a number of professional commitments he is passionate about. He is a full-time Senior Technology Advocate at PathAI, the Developer Community Advocate for the Julia Programming Language, and a Teaching Fellow for Harvard University's Extension School course CSCI E-33A.
Logan was previously an Applied Machine Learning Engineer and Software Engineer at Apple as well as the Community Manager for the Julia Programming Language. Additionally, Logan is on the Board of Directors at NumFOCUS and DEFNA. We started talking about the whole Julia Ecosystem with a particular focus on their pandemic response, touching a bit on the hot theme of the Metaverse. We spoke about why someone should use Julia with respect to other programming languages, mentioning some specific packages. We then switch to decentralisation/open source topics analyzing them ideologically, applicationally and financially. We then talked about Julia's future and the amazing interactions among Julia users. Given the background of Logan, we finally spoke about open science with NASA and the application of Julia in the aerospace sector, speaking also about PathAI, Logan's full-time company job.LINKS:
https://julialang.org
https://github.com/logankilpatrick
https://twitter.com/OfficialLoganK
https://scholar.harvard.edu/logankilpatrickRESOURCES:
Anchor: https://anchor.fm/poincare-podcast
Youtube: https://www.youtube.com/watch.v
RSS: https://anchor.fm/s/84561ce0/podcast/rss
Linktree: https://linktr.ee/poincaretrajectories
Company: https://www.linkedin.com/company/poincaretrajectories/ -
Here is a conversation with Erik Van Winkle, Operations Lead at DeSci Labs, besides having been part of the core team at Constitution DAO.
DeSci is a pioneering movement dedicated to exploring the capabilities of web3 technologies applied to the scientific ecosystem. We started talking about how DeSci works and the potential of sharing information while maintaining copyright. Then we jumped into the differences between decentralized science and open source, touching on the system design and business models. We then spoke about the story of DeSci and its relevant implementation choices. We finally discussed the new paradigm brought by DeSci, focusing on the ethics of interacting with traditional scientific media.LINKS:
https://www.linkedin.com/in/erik-van-winkle/
https://desci.com/
https://discord.gg/TnTsAUUu
https://t.me/BlockchainForScienceRESOURCES:
Anchor: https://anchor.fm/poincare-podcast
Youtube: https://www.youtube.com/watch.v
RSS: https://anchor.fm/s/84561ce0/podcast/rss
Linktree: https://linktr.ee/poincaretrajectories
Company: https://www.linkedin.com/company/poincaretrajectories/ -
Saknas det avsnitt?
-
The guest of this episode is Jean-Marc Mercier.Dr. Jean-Marc studies machine learning, both kernel methods and deep learning, in the context of mathematical finance.
We start talking about the differences between kernel methods and deep learning and some history of machine learning, then about the relations between orthogonal polynomials, and deep learning and kernel methods, touching on the application of kernel principal component analysis in aerospace and optimal transport. Dealing with finance, we talk about his vision in AI algorithmic trading and in general more financial applications where AI can be useful. Then we move on modelling approach and assumptions of the observable that brought us to economic bubble formation. We reserve quite a lot of time to talk about "codpy" an open-source python library for machine learning, mathematical finance and statistics of which Jean-Marc is one of the authors. We end up speaking about "codpy" more in detail such as function representation, mesh free methods which bring us to its applicability in fluid dynamics and we conclude with the future expansions of this library.LINKS:
https://www.researchgate.net/profile/Jean-Marc-Mercier
https://pypi.org/project/codpy/RESOURCES:
Anchor: https://anchor.fm/poincare-podcast
Youtube: https://www.youtube.com/watch.v
RSS: https://anchor.fm/s/84561ce0/podcast/rss
Linktree: https://linktr.ee/poincaretrajectories
Company: https://www.linkedin.com/company/poincaretrajectories/ -
Joel Rosenfeld is Assistant Professor at the University of South Florida and his research covers machine learning, kernel methods, approximation theory, function analysis and many more.
After a brief introduction about career strategies and issues caused by the pandemic, we talked about kernel functions with many digressions, such as the statistical description of dynamical systems and relationships between different spaces. We discuss the occupation of kernel functions and occupation measures (a branch of control theory) then we go on approximation of bounded operator using the densely defined operator and alternative approaches when operators are discontinuous in the domain. In the end, you'll find that sometimes one tries to make links between those subjects and dynamical systems and chaos, not all of them are congruent but is really interesting to hear the reasons.LINKS:
https://scholar.google.com/citations?user=pqsepdcAAAAJ&hl=en
https://www.thelearningdock.org
https://youtube.com/c/ThatMathThingRESOURCES:
Anchor: https://anchor.fm/poincare-podcast
Youtube: https://www.youtube.com/watch.v
RSS: https://anchor.fm/s/84561ce0/podcast/rss
Linktree: https://linktr.ee/poincaretrajectories
Company: https://www.linkedin.com/company/poincaretrajectories/ -
Gary Froyland is a professor at the School of Mathematics and Statistics at the University of New South Wales, and his research includes dynamical systems and optimization.
We started talking about research life and the struggle with our limits to pursue it, and then we went technical arguing about applied mathematics. In particular, we touched on the chaotic nature of climate and ocean science and the determination of some geometric structures (coherent) in it. We moved on to linear operators in functional analysis for time-varying dynamical systems and transfer operators both in discrete and continuous time. We discussed the software he developed (GAIO) and its applicability to the aerospace sector concluding by discussing the applicability of chaotic maps (e.g. Anosov), the periodicity of a system and the usage of the operators in multiscale systems (e.g. deterministic chaos and turbulence, fractals structures).LINKS:
https://scholar.google.com/citations?user=uAIy_MMAAAAJ&hl=en
https://web.maths.unsw.edu.au/~froyland/
https://research.unsw.edu.au/people/professor-gary-froylandRESOURCES:
Anchor: https://anchor.fm/poincare-podcast
Youtube: https://www.youtube.com/watch.v
RSS: https://anchor.fm/s/84561ce0/podcast/rss
Linktree: https://linktr.ee/poincaretrajectories
Company: https://www.linkedin.com/company/poincaretrajectories/ -
Miles Cranmer is a Ph.D. candidate at the University of Princeton. He is working on the interplay between astrophysics and AI.
We talk about Symbolic Regression, Genetic Programming, and the peculiarities of Julia, the programming language, comparing it with C++ and Python.
We then talk about his approach in studying new programming language features and about how to balance exploration versus exploitation. We largely discuss his work at Deepmind, outlining graph- and Lagrangian-Neural Networks, particularly in relation to the ability to investigate chaotic motion in dynamical systems.LINKS:
https://astroautomata.com
https://web.astro.princeton.edu/people/miles-cranmer
https://github.com/MilesCranmerRESOURCES:
Anchor: https://anchor.fm/poincare-podcast
Youtube: https://www.youtube.com/watch.v
RSS: https://anchor.fm/s/84561ce0/podcast/rss
Linktree: https://linktr.ee/poincaretrajectories
Company: https://www.linkedin.com/company/poincaretrajectories/ -
J. Nathan Kutz is the Robert Bolles and Yasuko Endo Professor within the Department of Applied Mathematics at the University of Washington in Seattle. He is also the director of the UW-led AI Institute for Dynamic Systems, founded by the National Science Foundation.
His research interests are numerical methods and scientific computing, data analysis and dimensionality reduction methods, dynamical systems, bifurcation theory, linear and nonlinear wave propagation, perturbation and asymptotic methods, nonlinear analysis, variational methods, soliton theory, nonlinear optics, mode-locked lasers, fluid dynamics, Bose-Einstein condensation, neuroscience, gesture recognition, and video & image processing.LINKS:
http://faculty.washington.edu/kutz/
https://www.youtube.com/channel/UCoUOaSVYkTV6W4uLvxvgiFA
J. Nathan Kutz Google ScholarRESOURCES:
Anchor: https://anchor.fm/poincare-podcast
Youtube: https://www.youtube.com/watch.v
RSS: https://anchor.fm/s/84561ce0/podcast/rss
Linktree: https://linktr.ee/poincaretrajectories
Company: https://www.linkedin.com/company/poincaretrajectories/ -
Igor Mezić is a mechanical engineer, mathematician, and Distinguished Professor of mechanical engineering and mathematics at the University of California, Santa Barbara. He is best known for his contributions to operator theoretic, data-driven approach to dynamical systems theory that he advanced via articles based on Koopman operator theory and his work on the theory of mixing.
LINKS:
Igor Mezic wiki
Igor Mezic LinkedIn
Igor Mezic ResearchGate ProfileRESOURCES:
Anchor: https://anchor.fm/poincare-podcast
Youtube: https://www.youtube.com/watch.v
RSS: https://anchor.fm/s/84561ce0/podcast/rss
Linktree: https://linktr.ee/poincaretrajectories
Company: https://www.linkedin.com/company/poincaretrajectories/ -
Emmanuel Blazquez is a Postdoctoral Research Fellow in Advanced Mission Analysis at the Advanced Concepts Team of the European Space Agency in Noordwijk, Nederland. His research areas are advanced mission analysis studies, with a focus on on-board real-time optimization asssisted by Artificial Intelligence. He also works on Autonomous Guidance and Control architectures, Multibody Astrodynamics and System Identification.
We start talking about the structure and goals of the Advanced Concepts Team, its interactions with ESA and the overall space community. We talk, as it is often the case in this podcast, about the synergies between Dynamical Systems and Artificial Intelligence. In particular, we talk about the use of classification techniques to investigate chaos, focusing in particular on Poincare' maps, starting from the relevance of chaotic dynamics in trajectory design.
We talk about potential interplay between convexification methods and Artificial Intelligence, in the context of optimal control and more in particular for randevouz, and how this has implications for the use of the Lunar Gateway.
We also talk about the use of GPUs in space research, talking about the asteroid aggregation problem, and the hardware he is using to conduct his work now.LINKS:
Emmanuel Blazquez LinkedIn
https://www.esa.int/gsp/ACT/team/emmanuel_blazquez/
Emmanuel Blazquez ResearchGate ProfileRESOURCES:
Anchor: https://anchor.fm/poincare-podcast
Youtube: https://www.youtube.com/watch.v
RSS: https://anchor.fm/s/84561ce0/podcast/rss
Linktree: https://linktr.ee/poincaretrajectories
Company: https://www.linkedin.com/company/poincaretrajectories/ -
Richard Linares is an assistant professor at MIT’s Department of Aeronautics and Astronautics and is the Co-Director of the Space Systems Laboratory. His research areas are astrodynamics, estimation and controls, satellite guidance and navigation, space situational awareness, and space traffic management.
We start talking about his interest in space traffic management, focusing on the importance of modelling the space environment: the thermosphere, the ionosphere and space weather. We discuss some of his works looking into reduced-order modelling techniques, like principal component analysis and dynamic mode decomposition, for the modelling of the thermospheric density field. We then discuss the use of machine learning techniques, like autoencoders and neural networks more in general, as promising generalizations, without neglecting their downsides.
Discussing the use of the Koopman Operator Theory in the same context, we move to its relevance in low dimensional, highly nonlinear dynamical systems, encountered every day in astrodynamics. We talk about its use for the study of the earth gravity field and for the construction of halo orbits in the restricted three-body problem. We discuss its implications for the engineering community, talking about optimal control, estimation and uncertainty quantification, about which we also outline the unification potential of techniques such as polynomial chaos expansion, differential algebra.
We then look into the potential of the Koopman Operator for dynamical systems theory in space. We discuss its potential for the analysis of invariant manifolds, its limitations for the study of chaotic systems, and finally its relations to the hamiltonian formalism of classical mechanics.LINKS:
http://arclab.mit.edu/
https://aeroastro.mit.edu/people/richard-linares/
Richard Linares LinkedInRESOURCES:
Anchor: https://anchor.fm/poincare-podcast
Youtube: https://www.youtube.com/watch.v
RSS: https://anchor.fm/s/84561ce0/podcast/rss
Linktree: https://linktr.ee/poincaretrajectories
Company: https://www.linkedin.com/company/poincaretrajectories/ -
Marian Gidea is a mathematician, focusing on the interplay between topology and dynamical systems theory, with a particular focus on Symplectic geometry and Hamiltonian dynamical systems. He is a Professor of Mathematics at Yeshiva University and Program Director at the National Science Foundation.
We start talking about practical applications of his work in celestial mechanics, in the context of the Lucy mission. We then get into the geometric properties of hamiltonian systems, from the foliation of invariant tori in the phase space, coming from the Liouville-Arnold theorem to the KAM theory, and the role of resonances for the arising of irregular motion. We discuss Arnold diffusion in celestial mechanics, the stability of the solar system, and the presence of stochastic behavior in celestial mechanics, not necessarily related to the presence of underlying stochastic processes (therefore underlying the relation between chaos and stochasticity). We mention the fundamental problem of dynamics by Poincaré and his conjecture about the denseness of periodic orbits in the three-body problem. We then move into the use of topological and machine learning techniques to characterize critical transitions in climate and financial systems. We close on the use of the Melnikov method and on the dialogue between the analysis of complex systems and Hamiltonian ones.LINKS:
https://www.yu.edu/faculty/pages/gidea-marian
Marian Gidea LinkedIn
Marian Gidea Google ScholarRESOURCES:
Anchor: https://anchor.fm/poincare-podcast
Youtube: https://www.youtube.com/watch.v
RSS: https://anchor.fm/s/84561ce0/podcast/rss
Linktree: https://linktr.ee/poincaretrajectories
Company: https://www.linkedin.com/company/poincaretrajectories/ -
Michael Betancourt is an experimental physicist and applied statistician focusing on the broad-scale applicability of bayesian reasoning in science.
He works as an independent consultant, and he is known as one of the core developers of Stan, a probabilistic programming language for statistical inference.We start talking about his physics background and his passion for Bayesian reasoning. We then talk about the challenges and opportunities of conducting research outside academia, from the necessity of alternative means of funding to the intellectual freedom gained outside institutions, being able to make everything freely accessible for example. We then introduce Bayesian inference and the potential of the Stan language in a number of scientific fields. We go inside what Stan does, talking about Markov Chain Monte Carlo, and focusing in particular on Hamiltonian Monte Carlo and on the good properties of symplectic geometry. We talk about the relations with convexification methods and the limitation of the use of Hilbert space methods and Machine Learning techniques. We close briefly touching on the Poincaré recurrence theorem, Hamiltonian chaos, and Geometric Ergodicity.LINKS:
https://betanalpha.github.io/
https://www.patreon.com/betanalpha/
https://twitter.com/betanalpha/RESOURCES:
Anchor: https://anchor.fm/poincare-podcast
Youtube: https://www.youtube.com/watch.v
RSS: https://anchor.fm/s/84561ce0/podcast/rss
Linktree: https://linktr.ee/poincaretrajectories
Company: https://www.linkedin.com/company/poincaretrajectories/ -
Jean-Philippe Bouchaud, a statistical physicist, is a pioneer in econophysics, a research field applying theories and methods originally developed by physicists in order to solve problems in economics, usually those including uncertainty or stochastic processes and nonlinear dynamics. He is the co-founder and chairman of Capital Fund Management, a global asset management using quantitative and scientific approaches to financial markets to invest billions of dollars in a systematic way. He is also the Head of Research of CFM and professor at École Normale Supérieure.
We talk about how ideas in dynamical systems theory and complex systems theory, like the ones developed by the 2021 Physics Nobel prize Giorgio Parisi, but also by Michael Fisher and Benoit Mandelbrot, influenced him. We talk about fat tails, Levy flights, and their emergence in both physical and financial systems. We talk about diffusion phenomena, fractional Brownian motion, hyperchaos, the Hurst exponent, and their application in finance. We touch on the wisdom of crowds, the emergence of intelligence in complex systems, their relations with the efficient market hypothesis, and the limits of a Markovian modeling of the financial market.
We also try to inform policymaking, both aiming at an optimal level of inequality in society and dealing with systematic incentives to push against what Bret Weinstein calls the personal responsibility vortex, therefore criticizing the invisible hand idea by Adam Smith. We close with the use of Artificial Intelligence techniques in finance, focusing on the relation between Deep Learning, Kernel Methods, and Random Matrix Theory.LINKS:
Jean-Philippe Bouchaud wiki
Jean-Philippe Bouchaud ResearchGate Profile
https://www.cfm.fr/
Jean-Philippe Bouchaud Google ScholarRESOURCES:
Anchor: https://anchor.fm/poincare-podcast
Youtube: https://www.youtube.com/watch.v
RSS: https://anchor.fm/s/84561ce0/podcast/rss
Linktree: https://linktr.ee/poincaretrajectories
Company: https://www.linkedin.com/company/poincaretrajectories/ -
Rasit Abay is an astrodynamicist and Machine Learning engineer. He is the founder of FuturifAI, a company whose aim is to accelerate AI adoption and to bring autonomy in use with Space and Earth data. He is focused on tackling challenges associated with Space Situational Awareness (SSA), like sensor fusion, collision avoidance, and maneuver optimization.
We talk about anti-satellite tests and the space debris issue, outlining the potential of a responsible space traffic management. We discuss some challenges he participated in, like to Collision Avoidance Challenge by ESA, and how his mixed background served him well in these scenarios. We touch on the use of reinforcement learning for space engineering problems, and how to view dynamical systems theory, like the Kolmogorov Arnold Moser theorem, using the tools of Artificial Intelligence.LINKS:
Rasit Abay LinkedIn
Rasit Abay ResearchGate Profile
https://futurifai.com/RESOURCES:
Anchor: https://anchor.fm/poincare-podcast
Youtube: https://www.youtube.com/watch.v
RSS: https://anchor.fm/s/84561ce0/podcast/rss
Linktree: https://linktr.ee/poincaretrajectories
Company: https://www.linkedin.com/company/poincaretrajectories/ -
Kevin Cowan is a professor at the Delft University of Technology. He graduated in aerospace engineering at the University of Texas at Austin and Delft University of Technology, with a focus on spaceflight and orbital mechanics.
He then worked as a research engineer at Applied Research Laboratories of the University of Texas at Austin. Subsequently, he pursued an MBA in international management at the Thunderbird School of Global Management. We talk about how his particular background is serving him in his academic life, we touch on the nature of scientific research and the power and limitations of the scientific method, the responsibilities of academics in society, and finally the influence of Artificial Intelligence in Space Engineering.
This project was initiated in the context of the Stardust Reloaded H2020 research network.LINKS:
https://online-learning.tudelft.nl/instructors/kevin-cowan/
Kevin Cowan ResearchGate ProfileCompany: https://www.linkedin.com/company/poincaretrajectories/
-
Daniel Pérez Palau is a professor at the Universidad Internacional de La Rioja.
He completed his PhD in Mathematics in the Dynamical Systems group at the University of Barcelona. After that, he joined the French Space Agency as a postdoctoral researcher, working on optimization problems. Now, together with teaching, he is focusing on new research areas, such as Artificial Intelligence.
This project was initiated in the context of the Stardust Reloaded H2020 research network.LINKS:
Daniel Pérez Palau Google ScholarCompany: https://www.linkedin.com/company/poincaretrajectories/
-
Angelo Cervone is Assistant Professor at TU Delft and responsible of several courses given by the Aerospace Engineering faculty: "Aerospace Design and Systems Engineering Elements", "Propulsion and Power", "Spacecraft Technology", "Micropropulsion".
He holds a PhD in space propulsion at University of Pisa (Italy), followed by a 2-years post-doc fellowship at Osaka University (Japan). He is author of more than 30 book chapters contributions and scientific articles, and more than 80 papers presented at international conferences. He is currently supervising 3 PhD candidates and has co-supervised 3 candidates who successfully completed their PhD in the past. He has been Project Manager or Principal Investigator for more than 10 projects, mainly funded by the European Space Agency. Among other ones, he is currently Principal Investigator at TU Delft for LUMIO, an international cooperation for a CubeSat mission at the Lagrangian point L2, to observe and characterize micrometeoroid impacts on the Lunar far side.
This project was initiated in the context of the Stardust Reloaded H2020 research network.LINKS:
Angelo Cervone ResearchGate Profile
https://online-learning.tudelft.nl/instructors/angelo-cervone/Company: https://www.linkedin.com/company/poincaretrajectories/
-
Elisabet Canalias is a mathematician and aerospace engineer working in the flight dynamics department of CNES, the French Space Agency. Her focus is on interplanetary trajectory design, and she is mostly known for her participation in the landing operations of the Philae probe on Comet 67P/Churyumov–Gerasimenko.
Her more recent endeavors, again focused on the exploration of minor bodies, include the Martian Moons eXporation mission and the trajectory design in strongly perturbed environments for the study of Phobos and Deimos and the Hayabusa2 mission, again in collaboration with the Japanese Aerospace Exploration Agency, particularly focused on the dynamics and operations of the MASCOT lander.
This project was initiated in the context of the Stardust Reloaded H2020 research network.LINKS:
Elisabet Canalias ResearchGate ProfileCompany: https://www.linkedin.com/company/poincaretrajectories/
-
Christopher Rackauckas is an Applied Mathematics Instructor at the Massachusetts Institute of Technology. He is also Senior Research Analyst at University of Maryland, Baltimore (School of Pharmacy) and the Director of Scientific Research at Pumas-AI and Creator / Lead Developer of Pumas.
He is arguably mostly known for being the Lead Developer of the SciML Open Source Software Organization and an active member of the Julia community, more in general.
This project was initiated in the context of the Stardust Reloaded H2020 research network.LINKS:
https://chrisrackauckas.com/index.html
https://sciml.ai/
https://diffeq.sciml.ai/dev/Company: https://www.linkedin.com/company/poincaretrajectories/
-
Edward Belbruno was once called in the New Scientist one of the "Top 10 most influential space thinkers". Indeed, Edward Belbruno is the man who revolutionized and is still revolutionizing our understanding of space travel. The work of Edward Belbruno shows us that methods from chaos theory of dynamical systems can be fruitfully applied to mission design. Charles Conley first proposed to take advantage of the dynamics of the restricted three-body problem to find low energy transit orbits from the Earth to the Moon, namely orbits which require much less fuel than the traditional orbits which are based on Keplerian motion. In 1986 Edward Belbruno found the first realistic low-energy transit orbit from an orbit around the Earth to an orbit around the Moon.
Apart from its practical relevance, the new discoveries of Edward Belbruno in capture dynamics lead as well to new insights in theoretical astrophysics. He discussed in the paper “Where Did The Moon Come From“ the origin of the Moon from the point of view of capture dynamics; a different application of Edward Belbruno’s Capture dynamics concerns the Lithopanspermia Hypothesis about the origin of life. Surprisingly, Edward Belbruno found in recent years quite striking relations between his previous works and different fields of physics: Black Holes, the Big Bang, quantum mechanics.
This project was initiated in the context of the Stardust Reloaded H2020 research network.LINKS:
https://www.edbelbruno.com/Company: https://www.linkedin.com/company/poincaretrajectories/
- Visa fler
