Forum Schedule Spring 2022
Most talks are expected to be conducted online
Fridays 3:45pm - 4:45pm BPB-217
Date | Speaker | Topic (click down-arrow to see abstract) | |
---|---|---|---|
Jan 21 |
Ming Zhong Texas A&M University, Institute of Data Science (TAMIDS) host: David Jeffery |
Machine Learning of Self-Organization from Observation |
|
Self-Organization can be used to explain crystal formation, aggregation of cells, social behaviors of insects, synchronization of heart beats, etc. It is challenging to understand these types of phenomena from the mathematical point of view. We offer a statistical/machine learning approach to understand these behaviors from observation; moreover, our learning approach can aid in validating and improving the modeling of Self-Organization. We develop a learning framework to derive physically meaningful dynamical models to explain the observation. We show the convergence property of our learning method in terms of the number of different initial conditions for first-order systems of homogeneous agents, and investigate its performance for various first- and second-order systems of heterogeneous agents. We also study the steady state properties of our learned models. We extend the learning approach to dynamical constrained on Riemannian manifolds, and we provide a convergence study for second order systems. Finally, we apply our learning method on the NASA Jet Propulsion Laboratory's modern Ephemerides. We find that our learned model can outperform the Newton's Universal Gravitation model in terms of reproducing the position/velocity of major celestial bodies, as well as preserving geometric properties (period/aphelion/perihelion) of the trajectory and highly-localized perihelion precession rates of Mars, Mercury, and the Moon. Upon careful inspection of our model, we discover that it even captures potion of the general relativity effects. Biography: Dr. Ming Zhong is currently a postdoctoral data scientist at Texas A&M Institute of Data Science (TAMIDS) in Texas A&M University. He works with the Scientific Machine Learning Lab at TAMIDS on developing algorithms for using machine learning methods (Deep Neural Networks and Gaussian Processes) to solve non-linear PDEs and their related inverse problems, with applications in radiative transfer equations, hyperbolic conservation laws, etc. Before his position at TAMIDS, he worked as a postdoctoral researcher at Johns Hopkins University with Mauro Maggioni on using machine learning methods to understand self-organization with applications in celestial mechanics. He obtained his Ph.D. from University of Maryland in Applied Mathematics under the guidance of Eitan Tadmor. |
|||
Jan 28 | |||
|
|||
Feb 4 | |||
|
|||
Feb 11 | |||
|
|||
Feb 18 | |||
|
|||
Feb 25 | |||
|
|||
Mar 4 | |||
|
|||
Mar 11 | |||
|
|||
Mar 18 | Spring Break | ||
|
|||
Mar 25 |
Chris White Princeton University host: Daniel Proga |
Black Hole Accretion: How to Connect Simulations to Observations and Learn Something in the Process |
|
There is a wealth of new and soon-to-be-made observations of material falling onto black holes, but how do we use images and light curves to learn about the underlying plasma, electromagnetic, and relativistic physics of these systems? I will discuss the use of fluid dynamics simulations for this purpose, including the importance of having the right analysis tools, covering parameter space, and understanding the physical processes at play. There will be particular emphasis on the rich dynamics that occurs when the accreting matter's angular momentum is misaligned with the spin of the black hole. |
|||
Apr 1 |
Melissa McClure Leiden University host: Zhaohuan Zhu |
A Song of Ice and Fire: Revealing the compositions of planetesimals in protoplanetary disks |
|
Volatile elements, like C, H, O, N, and S, are critical to the habitability of planets. The similarity between molecular isotopic ratios in Earth's oceans, comets, and protostars suggests that some of the volatile material in these bodies originated as ices in the dense molecular clouds from which stars are born. However, other ices, particularly the most volatile ones like CO, may be destroyed during the initial formation of a protoplanetary disk. Knowing which ices form in clouds, and survive to be incorporated into planetesimals, would be a huge step towards predicting whether exoplanets should have the ingredients to form life. In the first part of the talk, I will describe how upcoming Early Release Science observations with the successfully launched James Webb Space Telescope (JWST) will be used to reveal the ice and rock composition of solids in the planet-forming regions of protoplanetary disks. The question is then which of the available ices are actually incorporated into these planetesimals? To resolve this question, in the second part of the talk I demonstrate how the retention of ices in disks (either in planetesimals or millimeter sized pebbles) can be measured from infrared gas phase abundances in the hot inner 0.1 AU of protoplanetary disks. For an example disk, TW Hya, I show how the volatile abundances in the innermost disk imply the sequestration of C-rich ices in the disk's planet-formation zone. In combination with ALMA studies of ice traps in protoplanetary disks, this chemistry of cold ices and hot gas will provide a comprehensive record of which ices are (eventually) available to form life on other planets.
|
|||
Apr 8 |
Mohammad Safarzadeh NASA GSFC host: Bing Zhang |
The Astrophysical Context of Gravitational Wave Events |
|
We live in an era of breakthrough discoveries in gravitational waves (GW) astronomy. Such discoveries by LIGO/Virgo have been making headlines because the nature of these ev ents have been far from expectations. So, why are we puzzled? What is the road ahead for us to reach a deeper understanding? I focus on two of the most puzzling events: 1) The most massive binary black hole (BBH) merger with masses exceeding the pair-instability limit. I will present how a broader perspective on the host environment of BBHs can hold the key to understanding the nature of such systems. 2) The most massive binary neutron star merger. I will explore possible scenarios to explain first, why we have not detected such systems in the radio observations before. Secondly, how the key to understanding these events might lie in the r-process enrichment in the early universe and magnetic field evolution of neutron stars. Finally, I touch on how in the near future we will rely on artificial intelligence (AI)-driven discoveries in GW astrophysics, and how explainable-AI is vital for improving these models. |
|||
Apr 15 |
Richard Plotkin University of Nevada, Reno host: Jason Steffen |
Relativistic Jets from Weakly Accreting Black Holes |
|
The 'average' black hole in the local Universe accretes relatively weakly, emitting low levels of radiation in a 'quiescent’ spectral state. There is growing evidence that quiescent black holes always launch compact radio jets. However, there is still some uncertainty regarding how similar the physical properties of quiescent black hole jets are compared to compact jets launched by more rapidly accreting black holes. In this talk, I will present multiwavelength observations of stellar mass black holes (~10 Msun) in quiescent X-ray binary systems. I will discuss how our results are yielding new insights into how jets evolve as a function of accretion rate, and the implications for understanding jets from stellar mass, intermediate mass, and supermassive black holes in our Galaxy and beyond. |
|||
Apr 22 |
Susan Clark Stanford University host: Chao-Chin Yang |
The Magnetic Milky Way in Three Dimensions |
|
Magnetic fields thread our Milky Way Galaxy, influencing interstellar physics from cosmic ray propagation to star formation. The magnetic interstellar medium is also a formidable foreground for experimental cosmology, particularly for the quest to find signatures of inflation in the polarized cosmic microwave background (CMB). Despite its importance across scientific realms, the structure of the Galactic magnetic field is not well understood. Observational tracers like polarized dust emission yield only sky-projected, distance-integrated measurements of the three-dimensional magnetic structure. I will discuss new ways to probe interstellar magnetism in three dimensions, by combining high-resolution observations of Galactic neutral hydrogen with recent insights into how gas morphology encodes properties of the ambient magnetic field. These 3D maps are a new tool for understanding the magnetic interstellar medium and the polarized foreground to the CMB.
|
|||
Apr 29 |
Anna Ogorzalek NASA GSFC host: Daniel Proga |
Uncovering the Physics Behind AGN Feedback with High Resolution X-Ray Spectroscopy |
|
Active galactic nuclei (AGN) significantly impact the evolution of their host galaxies, as they can prevent star formation by either expelling large fractions of gas with wide-angle outflows, or heating up the halo gas with jets. However, how the AGN energy is transferred to the galaxy remains unknown. The X-ray band is key to answering this question, as the gas immediately impacted by the black hole reaches high, X-ray emitting temperatures. In this talk, I will present new applications of modern statistical techniques to high resolution X-ray spectra of nearby AGN in Seyferts, elliptical galaxies, and galaxy clusters. Here, using Bayesian approaches allows us to place competitive constraints on gas kinematics and thermodynamics, and gain new insights into the physical processes behind AGN feedback. I will conclude by introducing the Light Element Mapper, X-ray Probe mission concept, which will uncover the physics of the Circumgalactic Medium, the final frontier of AGN feedback and galaxy evolution. |
|||
May 6 |
Roman Krems University of British Columbia host: Bernard Zygelman |
Optimal quantum kernels for classification and regression problems |
|
I will describe our work that explores interplays of quantum physics and machine learning. The connection between quantum mechanics and machine learning is through kernels of reproducing kernel Hilbert spaces. I will describe an algorithm to construct kernels that yield Bayesian machine learning models capable of extrapolation in Hamiltonian parameter spaces. I will then show that this algorithm can be adapted for building optimal circuits of a gate-based quantum computer, yielding quantum kernels that outperform conventional classical kernels for small data machine learning tasks. If time permits, I will also show that support vector machines with a quantum kernel can be designed to be BQP-complete. References: Extrapolating quantum observables with machine learning: Inferring multiple phase transitions from properties of a single phase, R.A. Vargas-Hernandez, J. Sous, M. Berciu, and R.V. Krems, Physical Review Letters 121, 255702 (2018). Optimal quantum kernels for small data classification, E Torabian, RV Krems arXiv preprint arXiv:2203.13848 Quantum Gaussian process model of potential energy surface for a polyatomic molecule, J Dai, RV Krems arXiv preprint arXiv:2202.10601
|
|||
May 13 | Finals Week | ||
|
Past forums: Fall 2021 Spring 2021 Fall 2020 Spring 2020 Fall 2019 Spring 2019 Fall 2018 Spring 2018 Fall 2017 Spring 2017 Fall 2016 Spring 2016 Fall 2015 Spring 2015 Fall '14 Spring '14 Fall '13 Spring '13 Fall '12 Spring '12 Fall '11 Spring '11 Fall '10 Spring '10 Fall '09 Spring '09 Fall '08