BEGIN:VCALENDAR VERSION:2.0 PRODID:researchseminars.org CALSCALE:GREGORIAN X-WR-CALNAME:researchseminars.org BEGIN:VEVENT SUMMARY:YUEXIA LUNA LIN (Harvard University) DTSTART:20200710T160000Z DTEND:20200710T170000Z DTSTAMP:20260404T111101Z UID:CRIBB/1 DESCRIPTION:Title: Reference map technique: a fully Eulerian method for fluid-structure interactions\nby YUEXIA LUNA LIN (Harvard University) as part of Compu tational Research in Boston and Beyond Seminar (CRIBB)\n\n\nAbstract\nABST RACT:\n\nFluid-structure interactions (FSI) are abundantly observed in con texts ranging from swimming in the pool\, to industrial level manufacturin g\, to bacteria collective motion on a petri dish. However\, the governin g equations are only analytically trackable in the simple cases\, making s imulations key to understand this fantastic class of problems. Convention al computational methods often create a dilemma for fluid-structure intera ction (FSI) problems. Typically\, solids are simulated using a Lagrangian approach with a grid that moves with the material\, whereas fluids are si mulated using an Eulerian approach with a fixed spatial grid. FSI methods often require some type of interfacial coupling between the two different perspectives. We present a fully Eulerian FSI method that addresses these challenges. The method makes use of reference map\, which maps the solid in the current space to the reference space. Reference map is a common co ncept in finite strain theory\, but it has been under-utilized as a primar y variable for solid and FSI simulations. A challenge in applying the ref erence map technique (RMT) in FSI is to extrapolate reference map values f rom grid cells occupied by the solids to unoccupied grid cells\, in order to calculate derivative using finite difference schemes. This challenge b ecomes more acute when applying RMT to simulations in 3D. We develop an e xtrapolation algorithm based on least-square linear regression that is sui table for parallelization. We show examples to demonstrate that RMT is we ll suited for simulating soft\, highly-deformable materials and many-body contact problems. Joint work with Nicholas Derr and Chris H. Rycroft (SEA S\, Harvard University) and Ken Kamrin (Mechanical Engineering\, MIT).\n\n ZOOM:\n\nhttps://mit.zoom.us/j/96034732289\n Meeting ID: 960 3473 2289\n Password: 567284\n\n One tap mobile\n +1646 5588656\,\,96034732289# US (New York)\n +16699006833\,\,9603473228 9# US (San Jose)\n US : +1 646 558 8656 or +1 669 900 6833\n LOCATION:https://stable.researchseminars.org/talk/CRIBB/1/ END:VEVENT BEGIN:VEVENT SUMMARY:Jenelle Feather (MIT) DTSTART:20201002T160000Z DTEND:20201002T050000Z DTSTAMP:20260404T111101Z UID:CRIBB/2 DESCRIPTION:Title: Metamers of neural networks reveal divergence from human perceptual s ystems\nby Jenelle Feather (MIT) as part of Computational Research in Boston and Beyond Seminar (CRIBB)\n\n\nAbstract\nArtificial neural network s now achieve human-level performance on tasks such as image and speech re cognition\, raising the question of whether they should be taken seriously as models of biological sensory systems. Such neural network models exhib it human-like patterns of behavior\, and their feature spaces reliably pre dict brain activity. On the other hand\, neural network models can often b e fooled by small adversarial perturbations that have no effect on humans. In this talk\, I will detail our work using “model metamers” to inves tigate similarities between neural networks and human sensory systems. Mod el metamers are physically distinct stimuli that produce nearly the same r esponse within a model\, and thus the same model prediction. Our results s how that despite replicating aspects of human behavior and neural response s\, present-day deep neural networks learn invariances that deviate marked ly from those of biological sensory systems. Model metamers may help guide future model refinements to reduce or eliminate these discrepancies.\n\nh ttps://mit.zoom.us/j/96155042770 -- Meeting ID: 961 5504 2770\n LOCATION:https://stable.researchseminars.org/talk/CRIBB/2/ END:VEVENT BEGIN:VEVENT SUMMARY:Emily Crabb (MIT) DTSTART:20201106T170000Z DTEND:20201106T180000Z DTSTAMP:20260404T111101Z UID:CRIBB/3 DESCRIPTION:Title: Importance Of Equilibration Method and Sampling for Ab Initio Molecul ar Dynamics Simulations of Solvent - Lithium Salt Systems in Lithium-Oxyge n Batteries\nby Emily Crabb (MIT) as part of Computational Research in Boston and Beyond Seminar (CRIBB)\n\n\nAbstract\nZOOM Link: \n \ nhttps://mit.zoom.us/j/96155042770 - Meeting ID: 961 5504 2770\n\n====== =====================================================================\n\nL ithium-oxygen batteries are an active area of research because of their\np otential to have a much higher energy density than traditional lithium-ion \nbatteries. However\, they are not yet commercially viable due to poor\n efficiency\, high charging voltages\, and low cycle lifetimes. Many of th ese\nissues could be addressed with a deeper fundamental understanding of the\natomistic behavior of these batteries. One tool to model such atomic scale\nbehavior is ab initio molecular dynamics (AIMD) simulations. Howev er\, AIMD\nsimulations are limited to timescales of tens of picoseconds du e to their high\ncomputational cost. As a result\, equilibration and samp ling methodologies can\nhave a significant effect on the behavior of AIMD simulations. We thus\ncompared two equilibration methods for AIMD simulat ions of systems of common\nsolvents and salts found in lithium air batteri es: (1) using an AIMD\ntemperature ramp and (2) using a classical MD simul ation followed by a short\nAIMD simulation all at the target simulation te mperature of 300 K. We also\ncompared two different classical all-atom fo rce fields (PCFF+ and OPLS) and\nperformed multiple simulations for each s ystem. In this talk\, I will discuss\nwhy lithium-oxygen batteries are an exciting area of research\, why\ncomputational tools such as AIMD are cri tical to this field\, and how the\ndifferences between our simulation resu lts and experimental results for\nproperties such as coordination number i llustrate the importance of both\nequilibration method and independent sam pling for extracting experimentally\nrelevant quantities from AIMD simulat ions\, with applications in battery \ndevelopment and beyond.\n LOCATION:https://stable.researchseminars.org/talk/CRIBB/3/ END:VEVENT BEGIN:VEVENT SUMMARY:MIRIAM KREHER (MIT) DTSTART:20210205T170000Z DTEND:20210205T180000Z DTSTAMP:20260404T111101Z UID:CRIBB/4 DESCRIPTION:Title: Computational Analysis of Nuclear Reactor Transients\nby MIRIAM K REHER (MIT) as part of Computational Research in Boston and Beyond Seminar (CRIBB)\n\n\nAbstract\nZOOM MEETING info:\n\n h ttps://mit.zoom.us/j/96155042770\n\n Meetin g ID: 961 5504 2770\n\n\nSince nuclear experiments are costly and require extensive safety precautions\, the nuclear industry relies heavily on mode ling and simulation of nuclear systems. The state-of-the-art simulation t ool for steady-state neutron transport is Monte Carlo\, a probabilistic ap proach to solving for the distribution of neutrons. \n\nAlthough it is the most accurate tool available\, it is very computationally expensive. Mont e Carlo is even more burdensome when coupled to other physics which allows us to properly capture feedback effects from density and temperature chan ges. Nonetheless\, it is imperative to do such coupling because nuclear re actor designs rely on these intrinsic feedback mechanisms to ensure passiv e safety. In addition to coupling Monte Carlo with other physics codes\, t here is an additional hurdle to overcome for time-dependent simulations. These are a few of the reasons why nuclear reactor simulations are a targe t of Exascale computing initiatives. \n\nThis talk will cover a number of coupling schemes that create feasible runtimes for coupled time-dependent Monte Carlo simulations. In particular\, we will give consideration to hi gh-order/low-order schemes where Monte Carlo and diffusion solvers are pai red to deliver accurate results in efficient time. \n\n\nABOUT THE SPEAKER : Miriam Kreher is a PhD candidate in the Computational Reactor Physics G roup in the MIT Nuclear Science and Engineering Department. She is also a fellow of the DOE Computational Science Graduate Fellowship program. Kreh er received a BS in Engineering Science from the University of Pittsburgh in 2016. Kreher is a contributor of OpenMC and currently serves on the Boa rd of Directors of the American Nuclear Society.\n LOCATION:https://stable.researchseminars.org/talk/CRIBB/4/ END:VEVENT BEGIN:VEVENT SUMMARY:Kevin Silver (MIT) DTSTART:20210305T170000Z DTEND:20210305T180000Z DTSTAMP:20260404T111101Z UID:CRIBB/5 DESCRIPTION:Title: Buckling\, Crumpling\, And Tumbling Of Semiflexible Sheets In Simple Shear Flow\nby Kevin Silver (MIT) as part of Computational Research in Boston and Beyond Seminar (CRIBB)\n\n\nAbstract\nAs 2D materials such as graphene\, transition metal dichalcogenides\, and 2D polymers become more prevalent\, solution processing and colloidal-state properties are being e xploited to create advanced and functional materials. However\, our under standing of the fundamental behavior of 2D sheets and membranes in fluid f low is still lacking. In this work\, we perform numerical simulations of athermal semiflexible sheets with hydrodynamic interactions in shear flow. For sheets initially oriented in the flow-gradient plane\, we find buckl ing instabilities of different mode numbers that vary with bending stiffne ss and can be understood with a quasi-static model of elasticity. For diff erent initial orientations\, chaotic tumbling trajectories are observed.\n \n ZOOM MEETING info:\n\n https://mit.zoom.us/j/96155042770\n\n Meeting ID: 961 5504 2770\n LOCATION:https://stable.researchseminars.org/talk/CRIBB/5/ END:VEVENT BEGIN:VEVENT SUMMARY:Siddharth Samsi and Vijay Gadepally (MIT-Lincoln Lab) DTSTART:20210402T160000Z DTEND:20210402T170000Z DTSTAMP:20260404T111101Z UID:CRIBB/6 DESCRIPTION:Title: An Open Datacenter Dataset for AI Enabled Optimization\nby Siddha rth Samsi and Vijay Gadepally (MIT-Lincoln Lab) as part of Computational R esearch in Boston and Beyond Seminar (CRIBB)\n\n\nAbstract\nThe first step in training an AI is to get the right data. In order to apply AI to the problem of data center optimization\, such as identifying faults with serv ers\, energy or cooling systems\, before they become critical\, the MIT Li ncoln Laboratory Supercomputing Center is developing a state-of-the-art da taset. This dataset contains rich information such as: physical informati on about building management\; system information such as scheduler and fi lesystem logs\; and node-level information such as utilization\, memory\, GPU activity (both job level statistics as well as time-series monitoring collected via NVIDIA’s DCGM tool)\, energy utilization\, etc. In this ta lk\, we will describe the dataset\, detail how developers can get access t o this data\, and discuss a number of open problems associated with datace nter analytics.\n LOCATION:https://stable.researchseminars.org/talk/CRIBB/6/ END:VEVENT BEGIN:VEVENT SUMMARY:Peter James Ahrens (MIT) DTSTART:20210507T160000Z DTEND:20210507T170000Z DTSTAMP:20260404T111101Z UID:CRIBB/7 DESCRIPTION:Title: On Optimal Partitioning for Variable Block Row Format\nby Peter J ames Ahrens (MIT) as part of Computational Research in Boston and Beyond S eminar (CRIBB)\n\nLecture held in Virtual.\n\nAbstract\nThe Variable Block Row (VBR) format is an influential blocked sparse matrix format designed for matrices with shared sparsity structure between adjacent rows and colu mns. VBR groups adjacent rows and columns\, storing the resulting blocks t hat contain nonzeros in a dense format. This reduces the memory footprint and enables optimizations such as register blocking and instruction-level parallelism. Existing approaches use heuristics to determine which rows and columns should be grouped together. We show that finding the optimal grouping of rows and columns for VBR is NP-hard under several reasonable c ost models. In light of this finding\, we propose a 1-dimensional variant of VBR\, called 1D-VBR\, which achieves better performance than VBR by onl y grouping rows. We describe detailed cost models for runtime and memory consumption. Then\, we describe a linear time dynamic programming solutio n for optimally grouping the rows for 1D-VBR format. We extend our algorit hm to produce a heuristic VBR partitioner which alternates between optimal ly partitioning rows and columns\, assuming the columns or rows to be fixe d\, respectively. Our alternating heuristic produces VBR matrices with the smallest memory footprint of any partitioner we tested.\n\n ZOOM MEETING info:\n\n https://mit.zoom.us/j/96155042770\n\n Mee ting ID: 961 5504 2770\n LOCATION:https://stable.researchseminars.org/talk/CRIBB/7/ END:VEVENT BEGIN:VEVENT SUMMARY:Kurt Keville (MIT) DTSTART:20210604T160000Z DTEND:20210604T170000Z DTSTAMP:20260404T111101Z UID:CRIBB/8 DESCRIPTION:Title: Big Memory servers and modern approaches to disk-based computation\nby Kurt Keville (MIT) as part of Computational Research in Boston and B eyond Seminar (CRIBB)\n\n\nAbstract\nThere is a new computing paradigm ava ilable today facilitated by commodity server platforms. It is often called Big Memory solutions because it exposes a large RAM subsystem to the Oper ating System and therefore afford the application programmer a number of p reviously unavailable options for data management. Additionally\, certain vendor-specific solutions offer additional memory management options that pay dividends in data reliability and access speeds. A survey of these off erings and the promise of massive memory compute will be discussed.\n\n ZOOM MEETING info:\n\n https://mit.zoom.us/j/96155042770\n\n Meeting ID: 961 5504 2770\n LOCATION:https://stable.researchseminars.org/talk/CRIBB/8/ END:VEVENT BEGIN:VEVENT SUMMARY:Steven Torrisi (Harvard University) DTSTART:20210723T160000Z DTEND:20210723T170000Z DTSTAMP:20260404T111101Z UID:CRIBB/9 DESCRIPTION:Title: Which parts matter? Interpretable random forest models for X-Ray abs orption spectra\nby Steven Torrisi (Harvard University) as part of Com putational Research in Boston and Beyond Seminar (CRIBB)\n\n\nAbstract\nX- ray absorption spectroscopy (XAS) produces a wealth of information about t he local structure of materials\, but interpretation of spectra often reli es on easily accessible trends and prior assumptions about the structure. Recently\, researchers have demonstrated that machine learning models can automate this process to model the environments of absorbing atoms from th eir XAS spectra. However\, machine learning models are often difficult to interpret\, making it challenging to determine when they are valid and whe ther they are consistent with physical theories. In this work\, we present three main advances to the data-driven analysis of XAS spectra: we demons trate the efficacy of random forests in solving two new property determina tion tasks (predicting Bader charge and mean nearest neighbor distance)\, we address how choices in data representation affect model interpretabilit y and accuracy\, and we show that multiscale featurization can elucidate t he regions and trends in spectra that encode various local properties. The multiscale featurization transforms the spectrum into a vector of polynom ial-fit features\, and is contrasted with the commonly-used “pointwise ” featurization that directly uses the entire spectrum as input. We find that across thousands of transition metal oxide spectra\, the relative im portance of features describing the curvature of the spectrum can be local ized to individual energy ranges\, and we can separate the importance of c onstant\, linear\, quadratic\, and cubic trends\, as well as the white lin e energy. \n\nThis work has the potential to assist rigorous theoretical i nterpretations\, expedite experimental data collection\, and automate anal ysis of XAS spectra\, thus accelerating the discovery of new functional ma terials. We expect that this featurization strategy could be useful for br oad domains of application\, such as one-dimensional time-series analysis or other forms of spectroscopy.\n\nPaper: https://www.nature.com/articles/ s41524-020-00376-6\n\n==================================================== =======================\n\nZOOM MEETING info:\n\n https://mit. zoom.us/j/96155042770\n\n Meeting ID: 961 5504 2770\n LOCATION:https://stable.researchseminars.org/talk/CRIBB/9/ END:VEVENT BEGIN:VEVENT SUMMARY:Jenny Coulter (Harvard University) DTSTART:20210806T160000Z DTEND:20210806T170000Z DTSTAMP:20260404T111101Z UID:CRIBB/10 DESCRIPTION:Title: Phoebe: A New Open-Source Package for Electrical and Thermal Materia ls Transport Predictions from First-Principles\nby Jenny Coulter (Harv ard University) as part of Computational Research in Boston and Beyond Sem inar (CRIBB)\n\n\nAbstract\nUnderstanding the electrical and thermal trans port properties of materials is critical to the design of all kinds of dev ices. The theoretical prediction of these quantities relies on an accurate description of the electron and phonon properties of each material. Addit ionally\, a number of different quasiparticle interactions must be conside red to accurately predict transport behavior. While first-principles metho ds based on density functional theory can describe these material-specific quasiparticle properties\, using this information to calculate transport coefficients can be computationally demanding and memory intensive. \n\nTo address this challenge\, we present a recently developed software package \, Phoebe (https://github.com/mir-group/phoebe)\, which includes the effec ts of electron-phonon\, phonon-phonon\, boundary\, and isotope scattering to predict the electron and phonon transport properties of materials by so lving the Boltzmann transport equation (BTE) using a scattering matrix for malism. This open source C++ code utilizes MPI-OpenMP hybrid parallelizati on as well as GPU acceleration and distributed memory structures to manage computational cost and take advantage of modern HPC systems. Using this n ew framework\, we are able to accurately and efficiently predict a wide ra nge of material transport properties such as the electrical and thermal co nductivity and thermoelectric performance. \n\n https://math.mit.edu/sites/crib/\n\n ZOOM MEETING info:\n\n https://mit.zoo m.us/j/96155042770\n Meeting ID: 961 55 04 2770\n LOCATION:https://stable.researchseminars.org/talk/CRIBB/10/ END:VEVENT END:VCALENDAR