scholarly journals JAX-ReaxFF: A Gradient Based Framework for Extremely Fast Optimization of Reactive Force Fields

Author(s):  
Mehmet Cagri Kaymak ◽  
Ali Rahnamoun ◽  
Kurt A. O'Hearn ◽  
Adri C. T. van Duin ◽  
Kenneth M. Merz Jr. ◽  
...  

Molecular dynamics (MD) simulations facilitate the study of physical and chemical processes of interest. Traditional classical MD models lack reactivity to explore several important phenomena; while quantum mechanical (QM) models can be used for this purpose, they come with steep computational costs. The reactive force field (ReaxFF) model bridges the gap between these approaches by incorporating dynamic bonding and polarizability. To achieve realistic simulations using ReaxFF, model parameters must be optimized against high fidelity training data, typically with QM accuracy. Existing parameter optimization methods for ReaxFF consist of black-box techniques using genetic algorithms or Monte-Carlo methods. Due to the stochastic behavior of these methods, the optimization process can require millions of error evaluations for complex parameter fitting tasks, significantly hampering the rapid development of high quality parameter sets. In this work, we present JAX ReaxFF, a novel software tool that leverages modern machine learning infrastructure to enable extremely fast optimization of ReaxFF parameters. By calculating gradients of the loss function using the JAX library, we are able to utilize highly effective local optimization methods, such as the limited Broyden–Fletcher–Goldfarb–Shanno (LBFGS) and Sequential Least Squares Programming (SLSQP) methods. As a result of the performance portability of JAX, JAX-ReaxFF can execute efficiently on multi-core CPUs, GPUs (or even TPUs). By leveraging the gradient information and modern hardware accelerators, we are able to decrease parameter optimization time for ReaxFF from days to mere minutes. JAX-ReaxFF framework can also serve as a sandbox environment for domain scientists to explore customizing the ReaxFF functional form for more accurate modeling.

2020 ◽  
Vol 36 (9) ◽  
pp. 2690-2696
Author(s):  
Jarkko Toivonen ◽  
Pratyush K Das ◽  
Jussi Taipale ◽  
Esko Ukkonen

Abstract Motivation Position-specific probability matrices (PPMs, also called position-specific weight matrices) have been the dominating model for transcription factor (TF)-binding motifs in DNA. There is, however, increasing recent evidence of better performance of higher order models such as Markov models of order one, also called adjacent dinucleotide matrices (ADMs). ADMs can model dependencies between adjacent nucleotides, unlike PPMs. A modeling technique and software tool that would estimate such models simultaneously both for monomers and their dimers have been missing. Results We present an ADM-based mixture model for monomeric and dimeric TF-binding motifs and an expectation maximization algorithm MODER2 for learning such models from training data and seeds. The model is a mixture that includes monomers and dimers, built from the monomers, with a description of the dimeric structure (spacing, orientation). The technique is modular, meaning that the co-operative effect of dimerization is made explicit by evaluating the difference between expected and observed models. The model is validated using HT-SELEX and generated datasets, and by comparing to some earlier PPM and ADM techniques. The ADM models explain data slightly better than PPM models for 314 tested TFs (or their DNA-binding domains) from four families (bHLH, bZIP, ETS and Homeodomain), the ADM mixture models by MODER2 being the best on average. Availability and implementation Software implementation is available from https://github.com/jttoivon/moder2. Supplementary information Supplementary data are available at Bioinformatics online.


2019 ◽  
Vol 147 (5) ◽  
pp. 1429-1445 ◽  
Author(s):  
Yuchu Zhao ◽  
Zhengyu Liu ◽  
Fei Zheng ◽  
Yishuai Jin

Abstract We performed parameter estimation in the Zebiak–Cane model for the real-world scenario using the approach of ensemble Kalman filter (EnKF) data assimilation and the observational data of sea surface temperature and wind stress analyses. With real-world data assimilation in the coupled model, our study shows that model parameters converge toward stable values. Furthermore, the new parameters improve the real-world ENSO prediction skill, with the skill improved most by the parameter of the highest climate sensitivity (gam2), which controls the strength of anomalous upwelling advection term in the SST equation. The improved prediction skill is found to be contributed mainly by the improvement in the model dynamics, and second by the improvement in the initial field. Finally, geographic-dependent parameter optimization further improves the prediction skill across all the regions. Our study suggests that parameter optimization using ensemble data assimilation may provide an effective strategy to improve climate models and their real-world climate predictions in the future.


2021 ◽  
Vol 11 (9) ◽  
pp. 3827
Author(s):  
Blazej Nycz ◽  
Lukasz Malinski ◽  
Roman Przylucki

The article presents the results of multivariate calculations for the levitation metal melting system. The research had two main goals. The first goal of the multivariate calculations was to find the relationship between the basic electrical and geometric parameters of the selected calculation model and the maximum electromagnetic buoyancy force and the maximum power dissipated in the charge. The second goal was to find quasi-optimal conditions for levitation. The choice of the model with the highest melting efficiency is very important because electromagnetic levitation is essentially a low-efficiency process. Despite the low efficiency of this method, it is worth dealing with it because is one of the few methods that allow melting and obtaining alloys of refractory reactive metals. The research was limited to the analysis of the electromagnetic field modeled three-dimensionally. From among of 245 variants considered in the article, the most promising one was selected characterized by the highest efficiency. This variant will be a starting point for further work with the use of optimization methods.


2021 ◽  
Vol 10 (6) ◽  
pp. 420
Author(s):  
Jun Wang ◽  
Lili Jiang ◽  
Qingwen Qi ◽  
Yongji Wang

Image segmentation is of significance because it can provide objects that are the minimum analysis units for geographic object-based image analysis (GEOBIA). Most segmentation methods usually set parameters to identify geo-objects, and different parameter settings lead to different segmentation results; thus, parameter optimization is critical to obtain satisfactory segmentation results. Currently, many parameter optimization methods have been developed and successfully applied to the identification of single geo-objects. However, few studies have focused on the recognition of the union of different types of geo-objects (semantic geo-objects), such as a park. The recognition of semantic geo-objects is likely more crucial than that of single geo-objects because the former type of recognition is more correlated with the human perception. This paper proposes an approach to recognize semantic geo-objects. The key concept is that a single geo-object is the smallest component unit of a semantic geo-object, and semantic geo-objects are recognized by iteratively merging single geo-objects. Thus, the optimal scale of the semantic geo-objects is determined by iteratively recognizing the optimal scales of single geo-objects and using them as the initiation point of the reset scale parameter optimization interval. In this paper, we adopt the multiresolution segmentation (MRS) method to segment Gaofen-1 images and tested three scale parameter optimization methods to validate the proposed approach. The results show that the proposed approach can determine the scale parameters, which can produce semantic geo-objects.


Author(s):  
Huilin Zhou ◽  
Huimin Zheng ◽  
Qiegen Liu ◽  
Jian Liu ◽  
Yuhao Wang

Abstract Electromagnetic inverse-scattering problems (ISPs) are concerned with determining the properties of an unknown object using measured scattered fields. ISPs are often highly nonlinear, causing the problem to be very difficult to address. In addition, the reconstruction images of different optimization methods are distorted which leads to inaccurate reconstruction results. To alleviate these issues, we propose a new linear model solution of generative adversarial network-based (LM-GAN) inspired by generative adversarial networks (GAN). Two sub-networks are trained alternately in the adversarial framework. A linear deep iterative network as a generative network captures the spatial distribution of the data, and a discriminative network estimates the probability of a sample from the training data. Numerical results validate that LM-GAN has admirable fidelity and accuracy when reconstructing complex scatterers.


2009 ◽  
Vol 6 (4) ◽  
pp. 8279-8309 ◽  
Author(s):  
W. Ju ◽  
S. Wang ◽  
G. Yu ◽  
Y. Zhou ◽  
H. Wang

Abstract. Soil and atmospheric water deficits have significant influences on CO2 and energy exchanges between the atmosphere and terrestrial ecosystems. Model parameterization significantly affects the ability of a model to simulate carbon, water, and energy fluxes. In this study, an ensemble Kalman filter (EnKF) and observations of gross primary productivity (GPP) and latent heat (LE) fluxes were used to optimize model parameters significantly affecting the calculation of these fluxes for a subtropical coniferous plantation in southeastern China. The optimized parameters include the maximum carboxylation rate (Vcmax), the Ball-Berry coefficient (m) and the coefficient determining the sensitivity of stomatal conductance to atmospheric water vapor deficit D0). Optimized Vcmax and m showed larger seasonal and interannual variations than D0. Seasonal variations of Vcmax and m are more pronounced than the interannual variations. Vcmax and m are associated with soil water content (SWC). During dry periods, SWC at the 20 cm depth can explain 61% and 64% of variations of Vcmax and m, respectively. EnKF parameter optimization improves the simulations of GPP, LE and sensible heat (SH), mainly during dry periods. After parameter optimization using EnKF, the variations of GPP, LE and SH explained by the model increased by 1% to 4% at half-hourly steps and by 3% to 5% at daily time steps. Efforts are needed to develop algorithms that can properly describe the variations of these parameters under different environmental conditions.


2016 ◽  
Vol 9 (3) ◽  
pp. 118-137
Author(s):  
L.S. Kuravsky ◽  
P.A. Marmalyuk ◽  
G.A. Yuryev ◽  
O.B. Belyaeva ◽  
O.Yu. Prokopieva

This paper describes a new concept of flight crew assessment based on flight simulators training result. It is based on representation of pilot gaze movement with the aid of continuous-time Markov processes with discrete states. Considered are both the procedure of model parameters identification provided with goodness-of-fit tests in use and the classifier-building technique, which makes it possible to estimate degree of correspondence between the observed gaze motion distribution under study and reference distributions identified for different diagnosed groups. The final assessing criterion is formed on the basis of integrated diagnostic parameters, which are determined by the parameters of the identified models. The article provides a description of the experiment, illustrations, and results of studies aimed at assessing the reliability of the developed models and criteria, as well as conclusions about the applicability of the approach, its advantages and disadvantages.


Sign in / Sign up

Export Citation Format

Share Document