Showing 1–7 of 7 results for author: Strother, T

Capabilities of Gemini Models in Medicine

Authors: Khaled Saab, Tao Tu, Wei-Hung Weng, Ryutaro Tanno, David Stutz, Ellery Wulczyn, Fan Zhang, Tim Strother, Chunjong Park, Elahe Vedadi, Juanma Zambrano Chaves, Szu-Yeu Hu, Mike Schaekermann, Aishwarya Kamath, Yong Cheng, David G. T. Barrett, Cathy Cheung, Basil Mustafa, Anil Palepu, Daniel McDuff, Le Hou, Tomer Golany, Luyang Liu, Jean-baptiste Alayrac, Neil Houlsby , et al. (42 additional authors not shown)

Abstract: Excellence in a wide variety of medical applications poses considerable challenges for AI, requiring advanced reasoning, access to up-to-date medical knowledge and understanding of complex multimodal data. Gemini models, with strong general capabilities in multimodal and long-context reasoning, offer exciting possibilities in medicine. Building on these core strengths of Gemini, we introduce Med-G… ▽ More Excellence in a wide variety of medical applications poses considerable challenges for AI, requiring advanced reasoning, access to up-to-date medical knowledge and understanding of complex multimodal data. Gemini models, with strong general capabilities in multimodal and long-context reasoning, offer exciting possibilities in medicine. Building on these core strengths of Gemini, we introduce Med-Gemini, a family of highly capable multimodal models that are specialized in medicine with the ability to seamlessly use web search, and that can be efficiently tailored to novel modalities using custom encoders. We evaluate Med-Gemini on 14 medical benchmarks, establishing new state-of-the-art (SoTA) performance on 10 of them, and surpass the GPT-4 model family on every benchmark where a direct comparison is viable, often by a wide margin. On the popular MedQA (USMLE) benchmark, our best-performing Med-Gemini model achieves SoTA performance of 91.1% accuracy, using a novel uncertainty-guided search strategy. On 7 multimodal benchmarks including NEJM Image Challenges and MMMU (health & medicine), Med-Gemini improves over GPT-4V by an average relative margin of 44.5%. We demonstrate the effectiveness of Med-Gemini's long-context capabilities through SoTA performance on a needle-in-a-haystack retrieval task from long de-identified health records and medical video question answering, surpassing prior bespoke methods using only in-context learning. Finally, Med-Gemini's performance suggests real-world utility by surpassing human experts on tasks such as medical text summarization, alongside demonstrations of promising potential for multimodal medical dialogue, medical research and education. Taken together, our results offer compelling evidence for Med-Gemini's potential, although further rigorous evaluation will be crucial before real-world deployment in this safety-critical domain. △ Less

Submitted 1 May, 2024; v1 submitted 29 April, 2024; originally announced April 2024.

2D Implosion Simulations with a Kinetic Particle Code

Authors: Irina Sagert, Wesley P. Even, Terrance T. Strother

Abstract: We perform two-dimensional (2D) implosion simulations using a Monte Carlo kinetic particle code. The paper is motivated by the importance of non-equilibrium effects in inertial confinement fusion (ICF) capsule implosions. These cannot be fully captured by hydrodynamic simulations while kinetic methods, as the one presented in this study, are able to describe continuum and rarefied regimes within o… ▽ More We perform two-dimensional (2D) implosion simulations using a Monte Carlo kinetic particle code. The paper is motivated by the importance of non-equilibrium effects in inertial confinement fusion (ICF) capsule implosions. These cannot be fully captured by hydrodynamic simulations while kinetic methods, as the one presented in this study, are able to describe continuum and rarefied regimes within one approach. In the past, our code has been verified via traditional shock wave and fluid instability simulations. In the present work, we focus on setups that are closer to applications in ICF. We perform simple 2D disk implosion simulations using one particle species. The obtained results are compared to simulations using the hydrodynamics code RAGE. In a first study, the implosions are powered by energy deposition in the outer layers of the disk. We test the impact of the particle mean-free-path and find that while the width of the implosion shock broadens, its location as a function of time remains very similar. In a second study, we focus on the formation of fluid instabilities from induced perturbations. We find good agreement with hydrodynamic studies regarding the location of the shock, the implosion dynamics and the formation of imposed fluid instabilities. Differences are due to the higher resolution of RAGE and statistical noise in the kinetic simulations. The current studies represent proofs-of-principle and serve as a stepping stone for more sophisticated ICF simulations in the future. △ Less

Submitted 30 May, 2017; v1 submitted 14 November, 2016; originally announced November 2016.

Comments: 16 pages, 20 figures. Modified manuscript to match the published version

Journal ref: Phys. Rev. E 95, 053206 (2017)

Knudsen number dependence of 2D single-mode Rayleigh-Taylor fluid instabilities

Authors: Irina Sagert, Jim Howell, Alec Staber, Terrance Strother, Dirk Colbry, Wolfgang Bauer

Abstract: We present a study of single-mode Rayleigh-Taylor instabilities (smRTI) with a modified Direct Simulation Monte Carlo (mDSMC) code in two dimensions. The mDSMC code is aimed to capture the dynamics of matter for a large range of Knudsen numbers within one approach. Our method combines the traditional Monte Carlo technique to efficiently propagate particles and the Point-of-Closest-Approach method… ▽ More We present a study of single-mode Rayleigh-Taylor instabilities (smRTI) with a modified Direct Simulation Monte Carlo (mDSMC) code in two dimensions. The mDSMC code is aimed to capture the dynamics of matter for a large range of Knudsen numbers within one approach. Our method combines the traditional Monte Carlo technique to efficiently propagate particles and the Point-of-Closest-Approach method for high spatial resolution. Simulations are performed using different particle mean-free-paths and we compare the results to linear theory predictions for the growth rate including diffusion and viscosity. We find good agreement between theoretical predictions and simulations and, at late times, observe the development of secondary instabilities, similar to hydrodynamic simulations and experiments. Large mean-free-paths favor particle diffusion, reduce the occurrence of secondary instabilities and approach the non-interacting gas limit. △ Less

Submitted 9 July, 2015; v1 submitted 23 October, 2014; originally announced October 2014.

Comments: 15 pages, 12 figures, 1 table

Kinetic Simulations of Rayleigh-Taylor Instabilities

Authors: Irina Sagert, Wolfgang Bauer, Dirk Colbry, Jim Howell, Alec Staber, Terrance Strother

Abstract: We report on an ongoing project to develop a large scale Direct Simulation Monte Carlo code. The code is primarily aimed towards applications in astrophysics such as simulations of core-collapse supernovae. It has been tested on shock wave phenomena in the continuum limit and for matter out of equilibrium. In the current work we focus on the study of fluid instabilities. Like shock waves these are… ▽ More We report on an ongoing project to develop a large scale Direct Simulation Monte Carlo code. The code is primarily aimed towards applications in astrophysics such as simulations of core-collapse supernovae. It has been tested on shock wave phenomena in the continuum limit and for matter out of equilibrium. In the current work we focus on the study of fluid instabilities. Like shock waves these are routinely used as test-cases for hydrodynamic codes and are discussed to play an important role in the explosion mechanism of core-collapse supernovae. As a first test we study the evolution of a single-mode Rayleigh-Taylor instability at the interface of a light and a heavy fluid in the presence of a gravitational acceleration. To suppress small-wavelength instabilities caused by the irregularity in the separation layer we use a large particle mean free path. The latter leads to the development of a diffusion layer as particles propagate from one fluid into the other. For small amplitudes, when the instability is in the linear regime, we compare its position and shape to the analytic prediction. Despite the broadening of the fluid interface we see a good agreement with the analytic solution. At later times we observe the development of a mushroom like shape caused by secondary Kelvin-Helmholtz instabilities as seen in hydrodynamic simulations and consistent with experimental observations. △ Less

Submitted 29 July, 2014; originally announced July 2014.

Comments: Conference proceeding for The 30th Winter Workshop on Nuclear Dynamics 2014, 5 pages, 2 figures

Journal ref: J. Phys.: Conf. Ser. 535 012032 (2014)

Parallelization of Kinetic Theory Simulations

Authors: Jim Howell, Wolfgang Bauer, Dirk Colbry, Rodney Pickett, Alec Staber, Irina Sagert, Terrance Strother

Abstract: Numerical studies of shock waves in large scale systems via kinetic simulations with millions of particles are too computationally demanding to be processed in serial. In this work we focus on optimizing the parallel performance of a kinetic Monte Carlo code for astrophysical simulations such as core-collapse supernovae. Our goal is to attain a flexible program that scales well with the architectu… ▽ More Numerical studies of shock waves in large scale systems via kinetic simulations with millions of particles are too computationally demanding to be processed in serial. In this work we focus on optimizing the parallel performance of a kinetic Monte Carlo code for astrophysical simulations such as core-collapse supernovae. Our goal is to attain a flexible program that scales well with the architecture of modern supercomputers. This approach requires a hybrid model of programming that combines a message passing interface (MPI) with a multithreading model (OpenMP) in C++. We report on our approach to implement the hybrid design into the kinetic code and show first results which demonstrate a significant gain in performance when many processors are applied. △ Less

Submitted 12 October, 2013; originally announced October 2013.

Comments: 10 pages, 3 figures, conference proceedings

Journal ref: Proceedings of Nuclear Physics: Presence and Future, FIAS Interdisciplinary Science Series, Vol. 2 (Springer Verlag, 2014) pp 183-192

Building a Hydrodynamics Code with Kinetic Theory

Authors: Irina Sagert, Wolfgang Bauer, Dirk Colbry, Rodney Pickett, Terrance Strother

Abstract: We report on the development of a test-particle based kinetic Monte Carlo code for large systems and its application to simulate matter in the continuum regime. Our code combines advantages of the Direct Simulation Monte Carlo and the Point-of-Closest-Approach methods to solve the collision integral of the Boltzmann equation. With that, we achieve a high spatial accuracy in simulations while maint… ▽ More We report on the development of a test-particle based kinetic Monte Carlo code for large systems and its application to simulate matter in the continuum regime. Our code combines advantages of the Direct Simulation Monte Carlo and the Point-of-Closest-Approach methods to solve the collision integral of the Boltzmann equation. With that, we achieve a high spatial accuracy in simulations while maintaining computational feasibility when applying a large number of test-particles. The hybrid setup of our approach allows us to study systems which move in and out of the hydrodynamic regime, with low and high particle densities. To demonstrate our code's ability to reproduce hydrodynamic behavior we perform shock wave simulations and focus here on the Sedov blast wave test. The blast wave problem describes the evolution of a spherical expanding shock front and is an important verification problem for codes which are applied in astrophysical simulation, especially for approaches which aim to study core-collapse supernovae. △ Less

Submitted 17 May, 2013; originally announced May 2013.

Comments: Proceedings for the Winter Workshop on Nuclear Dynamics 2013, 6 pages, 6 figures

Journal ref: J. Phys.: Conf. Ser. 458 012031, 2013

Hydrodynamic Shock Wave Studies within a Kinetic Monte Carlo Approach

Authors: Irina Sagert, Dirk Colbry, Terrance Strother, Rodney Pickett, Wolfgang Bauer

Abstract: Kinetic approaches are routinely employed to simulate the dynamics of systems that are too rarified to be described by the Navier-Stokes equations. However, generally they are far too computationally expensive to be applied for systems that are governed by continuum hydrodynamics. In this paper, we introduce a massively parallelized test-particle based kinetic Monte Carlo code that is capable of m… ▽ More Kinetic approaches are routinely employed to simulate the dynamics of systems that are too rarified to be described by the Navier-Stokes equations. However, generally they are far too computationally expensive to be applied for systems that are governed by continuum hydrodynamics. In this paper, we introduce a massively parallelized test-particle based kinetic Monte Carlo code that is capable of modeling the phase space evolution of an arbitrarily sized system that is free to move in and out of the continuum limit. Using particle mean free paths which are small with respect to the characteristic length scale of the simulated system, we retrieve continuum behavior, while non-equilibrium effects are observed when the mean free path is increased. To demonstrate the ability of our code to reproduce hydrodynamic solutions, we apply a test-suite of classic hydrodynamic shock problems. Simulations using tens of millions of test-particles are found to reproduce the analytical solutions well. △ Less

Submitted 17 April, 2014; v1 submitted 30 October, 2012; originally announced October 2012.

Comments: 26 pages, 18 figures; corrections to text, new shock wave test added

Journal ref: Journal of Computational Physics, Volume 266, 1 June 2014, Pages 191-213