Below is a fairly comprehensive list of presentations I’ve given over the years. Where possible, I’ve included the the presentation/poster.
The conference name is a link to the conference website (if it still exists). Some of these presentations were made with Prezi. A Prezi viewer can be downloaded here.
2020
“Frequency-Dependent Material Motion Benchmarks for Radiative Transfer“, with N.A. Gentile, M&C 2021 Rayleigh (submitted).
“Temperature Dependence of Low-Energy Electron Transport in Water” Notre Dame Radiation Laboratory Seminar, June 25, 2020.
“Data Science in the Wild” Applied Data Science Workshop, Universidad Adolfo Ibáñez, Viña Del Mar, Chile. January 2020.
2019
“That’s an Interesting Idea: Data driven models, compressed sensing, and other outré tools for nuclear applications.” Young Members Research Achievement Award Plenary, M&C 2019 Portland.
“Non-Equilibrium Radiative Transfer Solutions using a Two-Group Diffusion Model” M&C 2019 Portland.
“Data Assimilation for Fission Neutron Multiplicity Data” with Ben Whewell and Simon Bolding. M&C 2019 Portland.
“Acceleration of Source Iteration using the Dynamic Mode Decomposition” with Terry Haut. M&C 2019 Portland.
“A low-rank method for time-dependent transport calculations” with Zhuogang Peng and Martin Frank. M&C 2019 Portland. Conference Paper
2018
“Machine Learning to Estimate Dante Response for Hohlraum Design” NECDC poster October 2018. (LA-UR-18-29211)
“Time-Eigenvalue Estimation using the Dynamic Mode Decomposition” NECDC poster October 2018. (LA-UR-18-29217)
“Data-Driven Algorithms for Transport” Lawrence Livermore WCI/CASC seminar, June 26, 2018. Read more
In this talk I will talk about particle transport algorithms that use data to form approximate operators to either accelerate traditional calculations or understand the properties of a given system. Based on the dynamic mode decomposition (DMD) of solutions or approximate solutions to the transport equation we construct a low-rank approximate transport operator using a singular value decomposition (SVD). With this transport operator we can accelerate the iterative convergence of source iteration in transport calculations or compute time-eigenvalues for a system based on the modes present in the system using on the time history of the solution. Results presented indicate that transport calculations can be accelerated using only source iterations plus an SVD, and that convergence is nearly independent of the scattering ratio of the system. Additionally, time-eigenvalues for a subcritical system are able to be calculated. Finally, a low-rank approximation to time-dependent problems is introduced and shown how it can be used for transport.
“Tail Fragility as a Tool For Model Confidence” ECCM-ECFD 2018, Glasgow, Scotland, UK, June 2018. Read more
Most computational models, that is, the combination of a mathematical model and a discretization scheme, are judged based on their performance in nominal cases. In terms of numerical methods the measures used might be convergence rates, stability, etc., and in mathematical modeling performance may be based on the accuracy of the model in known physical regimes.
In this presentation we adapt a measure of fragility developed for financial risk by Taleb and Douady based on the sensitivity of the tails of a distribution to parameters. Our adaptation looks at the probability of failure rather than cost of failure required in financial models.
2017
“One-Dimensional Models for Time-Dependent Transport in Solid Cylinders” 25th International Conference on Transport Theory, Monterey, California, October 2017. Read more
We present a set of 1-D transport models for solid cylinders of material irradiated with particles on the axial ends. The models are based on 1-D models originally developed for evacuated ducts with reflecting walls. The goal of this work is to show that 1-D models can be used for time-dependent transport in solid cylinders. A future use of this approach will apply these models to the high-energy density physics problem of a Marshak wave propagating down a cylindrical foam or other experiments. The models we present use a Galerkin procedure to project the radial dependence of the 3-D transport equation onto an expansion in terms of polynomials. Results demonstrate that for steady state problems with low scattering ratios, a three-basis function expansion can adequately capture the 3-D solution as computed via Monte Carlo. Smaller number of basis functions did not result in adequate solutions. As the radius of the cylinder increases, the 1-D model is more effective. Results from time dependent problems indicate that the 1-D models move particles too fast down the cylinder at early times, but are accurate on the order of 10 or more mean-free times. Our results indicate that 1-D models may be effective in modeling 2-D Marshak waves, but further work is necessary to answer this question.
“Predictive Science in Inertial Confinement Fusion The symphony of machine learning, high-performance computing, data science, uncertainty quantification, and multiphysics simulation”, University of Notre Dame Seminar, March 2017. Read more
This lecture will present the problem of predicting the results of inertial confinement fusion experiments using simulations. The basics of inertially confined nuclear fusion will be presented as well as the challenges of simulating these high-energy density systems. Along the way some brief discussion will be given to the presenter’s work in the theory, simulation, and uncertainty quantification of these predictions. The results of a recent simulation campaign using thousands of simulations and machine learning techniques to understand fusion will be discussed. The result of this campaign has been the discovery, via artificial intelligence, of novel implosion mechanisms that could be the key to achieving controlled ignition.
“High Fidelity, Moment-Based Methods for Particle Transport: The confluence of PDEs, Optimization, and HPC“, V-MAD 7 Séptimo Encuentro en Aplicaciones de la Matemática Instituto de Matemáticas, Pontificia Universidad Católica de Valparaíso, January, 2017. Read more
The calculation of the transport of particles is important in many applications including rarefied gas dynamics, plasma physics, and the nuclear energy systems. In this talk I will motivate the choice of moment-based methods for solving particle transport problems, and discuss the difficulties such approaches have. To obtain physically-meaningful solutions the discretization of the original integro-differential equation can depend on the solution to an optimization problem. I will show how these optimization problems arise, what methods perform best in terms of cost and accuracy, and how these problems can be well-suited for high performance computing.
2016
“Compressed Sensing for Nuclear Engineering“, North Carolina State Departmental Seminar, December, 2016. Read more
Compressed sensing reconstructs data using a small number of measurements and solving an underdetermined linear system. In this talk I will discuss some ways this technology can be applied to the nuclear engineering problems of radiation detection and computational radiation transport. The basics of compressed sensing will be discussed, including the single pixel camera, and then shown how the methods can be used in nuclear problems. In radiation detection, I will discuss the design of a single-detector neutron imaging system. Such a system can obtain images using fast neutrons, and could be a candidate for interrogation methods. I will also show how compressed sensing ideas can be used to improve the accuracy and speed of Monte Carlo transport calculations by defining tallies that are analogous to the single-pixel camera. Finally, I will show results of how these ideas can detect the important uncertainties in a simulation and even improve the finite element method.
“Polynomial Chaos Expansions for Uncertainty Quantification”, AICES EU Regional School, RWTH Aachen University, November, 2016. Part 1. Part 2. Read more
In computational science and engineering one often deals with computer simulations where inputs to the calculation are uncertain. A natural question to ask is how uncertain the output of a simulation is given uncertainties in the inputs. In this lecture I will give cover the application of orthogonal expansions in probability space (also known as polynomial chaos expansions) to determine the distribution of quantities of interest from a numerical simulation. I will detail how to apply these methods to a variety of input uncertainty distributions, and give concrete examples for simple functions as well as non-trivial applications. The examples will also be an opportunity to point out to students the pitfalls and common mistakes that can be made applying these techniques. Finally, I will cover more advanced ideas such as sparse quadrature and regularized regression techniques to estimate expansion coefficients.
“PDT: A Deterministic Transport Code for High-Performance and High-Fidelity Calculations“, DTRA Staff Seminar, July 2016.
“High-Performance Computing and Uncertainty Quantification for Particle Transport Problems: Beyond Nuclear Energy“, Pontificia Universidad Católica de Santiago, April, 2016.
“Leveraging Data Analysis to Improve Simulations Plus Additional Musings on Exascale Simulation“, DOE Conference on Data Analysis, March, 2016.
“High Fidelity, Moment-Based Methods for Particle Transport: The confluence of PDEs, Optimization, and HPC“, Workshop on Numerical PDEs Monash University, February 2016.
2015
“Efficient Estimation of Second-Order Sensitivity Coefficients“, Idaho National Laboratory Seminar, July 2015. Read more
To estimate second-order sensitivity coefficients (i.e., quadratic and interaction terms) for a quantity of interest in a simulation requires potentially many forward simulations to estimate. In this talk we will present an approach to estimating these coefficients by casting the sensitivities as terms in a regression model. Using regularized regression techniques we are able to accurately estimate these coefficients with many fewer simulation runs than perturbation theory suggests. In particular we solve a Monte Carlo model for the multiplication factor of a TRIGA fuel pin with 23 uncertain parameters and seek to estimate the 276 second-order sensitivity coefficients. The methods we used to calculate these include standard and Bayesian versions of ridge regression, lasso regression, and the Dantzig selector. Our results indicate that these methods can be used to effectively screen for important variables and to estimate the magnitude of higher-order sensitivities.
“On Variable Selection And Effective Estimation Of Interactive And Quadratic Sensitivity Coefficients: A Collection Of Regularized Regression Techniques”, ANS MC2015 – Joint International Conference on Mathematics and Computation (M&C), Supercomputing in Nuclear Applications (SNA) and the Monte Carlo (MC) Method, Nashville, Tennessee, April 19-23, 2015 – with Weixiong Zheng
2014
“Analysis of Lagged Weight Windows for Implicit Monte Carlo Variance Reduction”, ANS Winter Meeting 2014 – with Jacob Landman and Jonathan Madsen – November 2014.
“Measuring Angular Discretization Error During TRT Simulations with Energy Collapsed, Single Group Calculations“, ANS Winter Meeting 2014 – with Alex Long – November 2014.
“The Drift Diffusion Limit of Thermal Neutrons: Theoretical and Numerical Results”, PHYSOR2014 -given by Pablo Vaquer – Kyoto, Japan, October 2014.
“Effective Physics-Based Uncertainty Quan- tification for ZrHx Thermal Neutron Scattering in TRIGA Reactors”, PHYSOR2014 -given by Weixiong Zheng – Kyoto, Japan, October 2014.
“Uncertainty quantification and predictive science for high-energy density radiative transfer using neutron experiments“, 11th World Congress on Computational Mechanics, Barcelona, Spain – with the CERT team – July 2014. Read more
In this talk I will discuss recent uncertainty quantification work taking place at the Center for Exascale Radiation Transport (CERT), a multi-institution Predictive Science Academic Alliance Program (PSAAP) center funded by the US Department of Energy. The mission of CERT is to develop numerical methods and uncertainty quantification approaches for particle transport problems that are suitable for exascale computing platforms, as well as to perform the relevant particle transport experiments. My specific focus for this talk will be the development of a surrogate physics model and a physics-based impurity model for graphite. Our target application is the transport of thermal x-rays in high-energy density laboratory physics (HEDLP) experiments such as those in the area of inertial confinement fusion. That is we wish to be able to perform transport calculations for these experiments and quantify the uncertainty in predictions due to errors and uncertainties in the transport calculation. Nevertheless, a typical experiment in the HEDLP regime has several different physical processes occurring simultaneously with strong nonlinear coupling so that extracting the contribution to overall uncertainties from a single physical phenomenon (say x-ray transport) can be impossible due to limitations on experimental diagnostics. We have developed a method based on surrogate physics to create experiments involving only particle transport that are relevant to HEDLP x-ray transport. The experiments involve neutron transport through graphite, and such experiments are well understood from a measurement and diagnostic viewpoint (such experiments have been done for at least 60 years). In order to use these well understood problems to enable understanding of x-ray transport calculations, we have developed a novel mapping from neutron transport to x-ray transport.
“Uncertainty Quantification and Predictive Science for High‐Energy Density Radiative Transfer using Neutron Experiments“, AWE Staff Seminar – Aldermaston, UK – June, 2014.
“Fully Implicit Filtered PN for High-Energy Density Thermal Radiation Transport using Linear Discontinuous Galerkin Finite Elements“, HEDLA 2014, May, 2014 – with Vincent Laboure and Cory D. Hauck. Read more
The solution of thermal radiation transport as part of radiation-hydrodynamics calculations is important in the simulation of astrophysical phenomenon as well as high-energy density physics applications such as inertial confinement fusion. In this work we present an implicit method for solving the spherical harmonics (PN ) equations of radiation transport using filtered expansions. Since the introduction of such filtered PN methods by McClarren and Hauck, these approaches have been successful in producing high fidelity solutions to difficult transport problems. Nevertheless, there has been almost no investigation of the how to robustly and efficiently solve these equations implicitly in time – implicit integration is necessary unless one wants to evolve the flow at the speed of light time scale. Implicit solvers also impact the choice of filter strength. In this paper we present results of implicit filtered PN radiation transport simulations and discuss preconditioning strategies as well as the effect of implicit time integration on the necessary filter strength. We compare the results to reference Monte-Carlo calculations for several standard test problems, including radiation transport in a laser-driven shock tube experiment.
“High Energy Density Radiative Transfer Benchmark Solutions via Heterogeneous Computing“, HEDLA 2014, May, 2014 – with Daniel Holladay and John Wohlbier. Read more
High energy density radiative transfer benchmark solutions are presented for 1-D slab, cylindrical, and spherical geometries using a three-temperature (electron, ion, and radiation) model. A transport model is used for the radiation, a conduction model is used for the electrons, and ion motion is assumed negligible. These benchmarks are useful in the verification and testing of simulation codes for laboratory astrophysics as well as high-energy density physics. The solutions require linearization of the coupled equations and are obtained via specific cubic functional forms (in temperatures) for the heat capacities and electron-ion coupling factor. Comparisons to existing radiative transfer codes are presented. These solutions are semi-analytic in that their exact forms can be written down, but 2-D integrals must be computed numerically for each point in space and time. These integrals are slowly convergent and so a numerical integration routine was developed in OpenCL to take advantage of the high throughput that heterogeneous computing offers. Although capable of running on any OpenCL device, the nature of numerical integration meant GPUs were an excellent choice.
“Radiative Shock Solutions with Multigroup Discrete-Ordinates Transport“, HEDLA 2014, May, 2014 – with A. Miguel Holgado and Jim M. Ferguson. Read more
This study presents semi-analytic solutions of planar radiative shock waves with a multigroup discrete-ordinates transport model. Multigroup is used to capture the frequency-dependence of the material opacity and discrete-ordinates transport is used to capture the angular dependence of the radiation field. Comparisons are made with the grey non-equilibrium diffusion solutions of Lowrie and Edwards. These solutions can be used to verify radiation-hydrodynamics codes. By applying a multigroup discretization of the frequency dependence of the radiative energy intensity, we obtain a shock profile that is more representative of those found in actual radiative shocks. Using a transport model has shown that the anisotropy of the radiation field can cause the radiation energy density to be nonmonotonic and exhibit a local maximum if a Zel’dovich spike exists near the shock discontinuity. Several opacity problems are tested with the multigroup transport model and the results are presented.
“Understanding the Uncertainties in Neutron Transport Simulations“, University of Michigan Nuclear Engineering Departmental Seminar, Jan. 2014.
2013
“Physics-Based Uncertainty Quantication for the ZrHx Thermal Scattering Law“, Transactions of the American Nuclear Society, Nov 2013 – with Weixiong Zheng. Abstract (PDF).
“Improved Convergence Rates In Implicit Monte Carlo Simulations Through Stratified Sampling“, Transactions of the American Nuclear Society, Nov 2013 – with Alex Long Abstract (PDF).
“The Asymptotic Drift-Diffusion Limit of Thermal Neutrons” – The 23nd International Conference on Transport Theory (ICTT-23) – Santa Fe, New Mexico – September 2013 – with Marvin L. Adams. Abstract (PDF)
“Improved Discrete Ordinates Solutions Using Angular Filtering” – The 23nd International Conference on Transport Theory (ICTT-23) – Santa Fe, New Mexico – September 2013 – with Yuri Ayzman. Abstract (PDF)
“Self-Similar Radiation-Hydrodynamics Solutions in the Equilibrium Diffusion Limit ” – Multimat 2013 – September 6, 2013 San Francisco, CA – with Taylor K. Lane. Read more
This work presents semi-analytic solutions to a radiation-hydrodynamics problem of a radiation source driving an initially cold medium. Our solutions are in the equilibrium diffusion limit, include material motion and allow for radiation-dominated situations where the radiation energy is comparable to (or greater than) the material internal energy density. As such, this work is a generalization of the classical Marshak wave problem that assumes no material motion and assumes that the radiation energy is negligible. The solutions provide insight into the impact of radiation energy and material motion, as well as present a novel verification test for radiation transport packages. These problems can be solved by the radiation package by prescribing the material velocity to have the appropriate behavior in time. The solution thereby tests the radiation-matter coupling terms and their v/c treatment without needing a hydrodynamics solve.
“Open Problems in the Preconditioning of Moment-based Transport Equations” – Struktura Katedry Matematiky, University of West Bohemia Seminar – June, 2013.
“Temperature-Extrapolation Method For Implicit Monte Carlo – Radiation Hydrodynamics Calculations“, International Conference on Mathematics and Computational Methods Applied to Nuclear Science & Engineering (M&C 2013) Sun Valley, Idaho, USA, May 5-9, 2013 – with Todd J. Urbatsch. Paper Read more
We present a method for implementing temperature extrapolation in Implicit Monte Carlo solutions to radiation hydrodynamics problems. The method is based on a BDF-2 type integration to estimate a change in material temperature over a time step. We present results for radiation only problems in an infinite medium and for a 2-D Cartesian hohlraum problem. Additionally, radiation hydrodynamics simulations are presented for an RZ hohlraum problem and a related 3D problem. Our results indicate that improvements in noise and general behavior are possible. We present considerations for future investigations and implementations.
2012
“An Implicit Monte Carlo Method Based on BDF-2 Time Integration for Simulating Nonlinear Radiative Transfer” – American Nuclear Society Winter Meeting 2012 – San Diego, CA – November 2012 – with Todd J. Urbatsch. Abstract (PDF)
“A Flux-Limited Diffusion Method for Simulating Radiative Shocks” – American Nuclear Society Winter Meeting 2012 – San Diego, CA – November 2012 – with Taylor K. Lane. Abstract (PDF)
“A BDF-2 Time Integration Method for Simulating Radiative Transfer (U)” – NECDC – Livermore, CA – October 2012 – with Todd J. Urbatsch
“How To Plan a Successful Big Data Pilot” Strata + Hadoop World, New York, October, 2012 – with Michael Gold. Read more
Big data initiatives often begin with a pilot project. This approach can generate internal support to invest in larger big data initiatives by demonstrating the positive impact of data science and data-driven decisions on desired business outcomes (e.g. profits, margins, etc.) and business processes (e.g. creating rubrics for making decisions and metrics to measure the impact) Nevertheless, executing pilot projects can be difficult, and many pilots don’t convert into larger big data projects.In this session we’ll explore the challenges of big data pilots farsite has encountered and suggest ways to plan and execute a success pilot.
“Using Mobile Technologies for Real Estate Research” ICSC Research Connections, Chicago, October, 2012 – with Michael Gold. Read more
In the rapidly changing world of mobile phones and mobile applications, a powerful new source of consumer information is emerging as a research tool. a panel of experts presents several studies on the use of mobile technologies and social media technologies for site location analysis, conducting market research and determining shopping movement and behavior. a discussion of the benefits, challenges, and privacy aspects of these new technologies will be presented.
“Bayesian MARS UQ Research at the Center for Radiative Shock Hydrodynamics (CRASH)” – Oak Ridge National Laboratory Seminar – September 2012 – with H.F. Stripling. Read more
I will talk about the experiments and predictions being performed at the Center for Radiative Shock Hydrodynamics. For these laser-driven shock experiments we have embarked on a campaign to predict future experiments. Part of our prediction uses small-scale experiments to calibrate our codes. Using Bayesian response surface techniques (also known as emulators) based on Bayesian Multiple Adaptive Regression Splines (BMARS), we have produced an approach to calibrate using experimental data under non-normal uncertainties. Additionally, I will describe an extension to BMARS that improves emulator accuracy when gradient information is available.
“Customer analytics in a multichannel world” – TDWI BI EXECUTIVE SUMMIT San Diego, CA, July 30–August 1, 2012 – with Michael Gold. Read more
Despite the fast-paced growth of online retailing, customers do not engage with companies through just one channel. consumers still go to stores to browse merchandise, discuss with knowledgeable salespeople, and experience the in-store environment. People like to go to stores; thus, Ffor customer intelligence, it is important to understand customer behavior across these multiple channelsin-store and online channels. Analyzing geospatial data is becoming increasingly vital to this analysis. Multi-channel data analysis is important for store and merchandise location, advertising, merchandising, and marketing decisions.
In this session, Michael Gold and Ryan McClarren of farsite will discuss how advanced data analysis, includingof geospatial analysis, can support a multi-channel approach to retail customer analytics. Key takeaways include:
- Integrating social media data and analysis into strategic decision making
- Linking online and offline consumer behavior
- Exploring the value of combining merchandising, marketing, and real estate data
- Discussing the importance of the location component of data in retail customer analytics
“VV/UQ Implications of Performance Models for Large Scale Computing” – 2012 PSAAP VV/UQ Workshop – Ann Arbor, MI – August 2012.
“Preconditioning Strategies for Particle Transport Simulations“, Seminar, Struktura Katedry Matematiky, University of West Bohemia Seminar – July 2012.
2011
“Gradient Enhanced Bayesian MARS for Regression and Uncertainty Quantification” – American Nuclear Society Winter Meeting 2011 – San Diego, CA – November 2012 – given by H.F. Stripling. Abstract (PDF)
“New Faculty Introduction NUEN Advisory Council Presentation“, Texas A&M Nuclear Engineering Department Advisory Council Meeting, October, 2011.
“A P2-Equivalent Form of the SP2 Equations” – The 22nd International Conference on Transport Theory (ICTT-22) – Portland, Oregon – September 2011. Abstract (PDF)
“Robust and Accurate Methods for Thermal Radiation Transport“, Earth Science and Engineering Colloquium, Imperial College, London, UK, August, 2011.
“Predicting Radiating Shock Experiments: Theory, VV/UQ, and more at CRASH“, AWE Staff Seminar – Aldermaston, UK – August, 2011.
“Uncertainty Quantification, Predictive Science, and Related Topics“, Multiphysics Methods Workshop, Sunriver, Oregon, June, 2010, with Marvin L. Adams.
“The Spectral Volume Method as Applied to Transport Problems“, International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering (M&C 2011) – Rio De Janeiro – May 2011.
“Bayesian MARS For Uncertainty Quantification In Stochastic Transport Problems“, International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering (M&C 2011) – Rio De Janeiro – May 2011 – given by Hayes F. Stripling.
“Calibration of Uncertain Inputs to Computer Models Using Experimentally Measured Quantities and the BMARS Emulator“, International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering (M&C 2011) – Rio De Janeiro – May 2011 – given by Hayes F. Stripling.