Search

Classical force fields (FFs) based on machine learning (ML) methods show great potential for large scale simulations of solids.

A general method to invert parameter distributions of a polydisperse system using data acquired from a small-angle scattering (SAS) experiment is presented. The forward problem, i.e. calculating the scattering intensity given the distributions of any causal parameters of a theoretical model, is generalized as a multi-linear map, characterized by a high-dimensional Green tensor that represents the complete scattering physics. The inverse problem, i.e. finding the maximum-likelihood estimation of the parameter distributions (in free form) given the scattering intensity (either a curve or an image) acquired from an experiment, is formulated as a constrained nonlinear programming (NLP) problem. This NLP problem is solved with high accuracy and efficiency via several theoretical and computational enhancements, such as an automatic data scaling for accuracy preservation and GPU acceleration for large-scale multi-parameter systems. Six numerical examples are presented, including both synthetic tests and solutions to real neutron and X-ray data sets, where the method is compared with several existing methods in terms of their generality, accuracy and computational cost. These examples show that SAS inversion is subject to a high degree of non-uniqueness of solution or structural ambiguity. With an ultra-high accuracy, the method can yield a series of near-optimal solutions that fit data to different acceptable levels.

The breakthrough in Deep Learning neural networks has transformed the use of AI and machine learning technologies for the analysis of very large experimental datasets. These datasets are typically generated by large-scale experimental facilities at national laboratories. In the context of science, scientific machine learning focuses on training machines to identify patterns, trends, and anomalies to extract meaningful scientific insights from such datasets. With a new generation of experimental facilities, the rate of data generation and the scale of data volumes will increasingly require the use of more automated data analysis. At present, identifying the most appropriate machine learning algorithm for the analysis of any given scientific dataset is still a challenge for scientists. This is due to many different machine learning frameworks, computer architectures, and machine learning models. Historically, for modelling and simulation on HPC systems such problems have been addressed through benchmarking computer applications, algorithms, and architectures. Extending such a benchmarking approach and identifying metrics for the application of machine learning methods to scientific datasets is a new challenge for both scientists and computer scientists. In this paper, we describe our approach to the development of scientific machine learning benchmarks and review other approaches to benchmarking scientific machine learning.

SUMMARYDistal Visceral Endoderm (DVE) cells show a stereotypic unidirectional migration essential for correct orientation of the anterior-posterior axis. They migrate within a simple epithelium, the Visceral Endoderm (VE). It is unknown how DVE cells negotiate their way amongst the surrounding VE cells, what determines the bounds of DVE migration within the VE, and the relative contributions of different cell behaviours to this migration. To address these questions, we used lightsheet microscopy to generate a multi-embryo, singlecell resolution, longitudinal dataset of cell behaviour and morphology. We developed a machine learning based pipeline to segment cells and a data-informed systematic computational framework to extract and compare select morphological, behavioural and molecular parameters of all VE cells in a unified coordinate space. Unbiased clustering of this single-cell ‘phenomic’ dataset reveals considerable patterned phenotypic heterogeneity within the VE and a previously unknown sub-grouping within the DVE. While migrating, DVE cells retain regular morphology, do not exchange neighbours and are crowded, all hallmarks of the jammed state. In contrast, VE cells immediately ahead of them deform and undergo neighbour exchange. We show that DVE cells are characterised by higher levels of apical F-actin and elevated tension relative to the VE cells immediately ahead of them through which they migrate, but stop migrating upon reaching a region of the VE with matching elevated tension.Lefty1mutants, known to show abnormal over-migration of DVE cells, show disruption to this patterned tension in the VE. Our findings provide novel insights into the control of cell behaviour during the remodelling of curved epithelia, indicating that the collective migration of sub-sets of cells can be circumscribed by modulating the mechanical properties of surrounding cells and that migrating cells in this context remain as a jammed solid flock, with surrounding cells facilitating their movement by becoming unjammed.Graphical Abstract

Knowing the redshift of galaxies is one of the first requirements of many cosmological experiments, and as it is impossible to perform spectroscopy for every galaxy being observed, photometric redshift (photo-z) estimations are still of particular interest. Here, we investigate different deep learning methods for obtaining photo-z estimates directly from images, comparing these with ‘traditional’ machine learning algorithms which make use of magnitudes retrieved through photometry. As well as testing a convolutional neural network (CNN) and inception-module CNN, we introduce a novel mixed-input model that allows for both images and magnitude data to be used in the same model as a way of further improving the estimated redshifts. We also perform benchmarking as a way of demonstrating the performance and scalability of the different algorithms. The data used in the study comes entirely from the Sloan Digital Sky Survey (SDSS) from which 1 million galaxies were used, each having 5-filtre (ugriz) images with complete photometry and a spectroscopic redshift which was taken as the ground truth. The mixed-input inception CNN achieved a mean squared error (MSE) =0.009, which was a significant improvement ($30{{\ \rm per\ cent}}$) over the traditional random forest (RF), and the model performed even better at lower redshifts achieving a MSE = 0.0007 (a $50{{\ \rm per\ cent}}$ improvement over the RF) in the range of z < 0.3. This method could be hugely beneficial to upcoming surveys, such as Euclid and the Vera C. Rubin Observatory’s Legacy Survey of Space and Time (LSST), which will require vast numbers of photo-z estimates produced as quickly and accurately as possible.

Supervised machine learning (ML) models are frequently trained on large datasets of physics-based simulations with the aim of being applied to experimental data. However, ML models trained on simulated data often struggle to perform on experimental data, because there is a shift in the data caused by experimental artefacts that might be challenging to simulate. We introduce Exp2SimGAN, an unsupervised image-to-image ML model to match simulated and experimental data. To train, Exp2SimGAN only requires a set of experimental data and a set of (not necessarily corresponding) simulated data. Once trained, it can convert a simulated dataset into one that resembles an experiment, and vice versa. We demonstrate that Exp2SimGAN can be used to match simulated and experimental two- and three-dimensional inelastic neutron scattering (INS) spectra, enabling the analysis of experimental INS data using supervised ML. Finally, we provide a domain of application measure for Exp2SimGAN, allowing us to assess the likelihood that Exp2SimGAN will be successful on a specific dataset. Exp2SimGAN is a step towards analysis of experimental data using supervised ML models trained on physics-based simulations.

We shed light on a pitfall and an opportunity in physics-informed neural networks (PINNs). We prove that a multilayer perceptron (MLP) only with ReLU (Rectified Linear Unit) or ReLU-like Lipschitz activation functions will always lead to a vanished Hessian. Such a network-imposed constraint contradicts any second- or higher-order partial differential equations (PDEs). Therefore, a ReLU-based MLP cannot form a permissible function space for the approximation of their solutions. Inspired by this pitfall, we prove that a linear PDE up to the $n$-th order can be strictly satisfied by an MLP with $C^n$ activation functions when the weights of its output layer lie on a certain hyperplane, as called the out-layer-hyperplane. An MLP equipped with the out-layer-hyperplane becomes "physics-enforced", no longer requiring a loss function for the PDE itself (but only those for the initial and boundary conditions). Such a hyperplane exists not only for MLPs but for any network architecture tailed by a fully-connected hidden layer. To our knowledge, this should be the first PINN architecture that enforces point-wise correctness of PDEs. We show a closed-form expression of the out-layer-hyperplane for second-order linear PDEs, which can be generalised to higher-order nonlinear PDEs., 12 pages, 3 figures

Keyhole porosity is a key concern in laser powder-bed fusion (LPBF), potentially impacting component fatigue life. However, some keyhole porosity formation mechanisms, e.g., keyhole fluctuation, collapse and bubble growth and shrinkage, remain unclear. Using synchrotron X-ray imaging we reveal keyhole and bubble behaviour, quantifying their formation dynamics. The findings support the hypotheses that: (i) keyhole porosity can initiate not only in unstable, but also in the transition keyhole regimes created by high laser power-velocity conditions, causing fast radial keyhole fluctuations (2.5–10 kHz); (ii) transition regime collapse tends to occur part way up the rear-wall; and (iii) immediately after keyhole collapse, bubbles undergo rapid growth due to pressure equilibration, then shrink due to metal-vapour condensation. Concurrent with condensation, hydrogen diffusion into the bubble slows the shrinkage and stabilises the bubble size. The keyhole fluctuation and bubble evolution mechanisms revealed here may guide the development of control systems for minimising porosity.

The Research Software Engineer Knowledge IntegrationLandscape Review was written by the ExCALIBUR Programme Community. It documents the concept of the Research Software Engineer (RSE) and their role in many of todays computational science projects.The reviewprovides a comprehensive review of the skills required by Research Software Engineersin HPC and their future training needs

Structures of macromolecular assemblies derived from cryo-EM maps often contain errors that become more abundant with decreasing resolution. Despite efforts in the cryo-EM community to develop metrics for map and atomistic model validation, thus far, no specific scoring metrics have been applied systematically to assess the interface between the assembly subunits. Here, we comprehensively assessed protein–protein interfaces in macromolecular assemblies derived by cryo-EM. To this end, we developed Protein Interface-score (PI-score), a density-independent machine learning-based metric, trained using the features of protein–protein interfaces in crystal structures. We evaluated 5873 interfaces in 1053 PDB-deposited cryo-EM models (including SARS-CoV-2 complexes), as well as the models submitted to CASP13 cryo-EM targets and the EM model challenge. We further inspected the interfaces associated with low-scores and found that some of those, especially in intermediate-to-low resolution (worse than 4 Å) structures, were not captured by density-based assessment scores. A combined score incorporating PI-score and fit-to-density score showed discriminatory power, allowing our method to provide a powerful complementary assessment tool for the ever-increasing number of complexes solved by cryo-EM., Here, the authors present Protein Interface-score (PI-score), a machine learning-based metric that has been trained on protein–protein interfaces’ features from high-resolution crystal structures. They use the PI-score to evaluate the protein–protein interfaces in more than 1000 PDB-deposited cryo-EM structures and show that it can be used as a complementary assessment tool for cryo-EM model validation.

Detection of multiple lane markings on road surfaces is an important aspect of autonomous vehicles. Although a number of approaches have been proposed to detect lanes, detecting multiple lane markings, particularly across a large number of frames and under varying lighting conditions, in a consistent manner is still a challenging problem. In this paper, we propose a novel approach for detecting multiple lanes across a large number of frames and under various lighting conditions. Instead of resorting to the conventional approach of processing each frame to detect lanes, we treat the overall problem as a multitarget tracking problem across space and time using the integrated probabilistic data association filter (IPDAF) as our basis filter. We use the intensity of the pixels as an augmented feature to correctly group multiple lane markings using the Hough transform. By representing these extracted lane markings as splines, we then identify a set of control points, which becomes a set of targets to be tracked over a period of time, and thus across a large number of frames. We evaluate our approach on two different fronts, covering both model- and machine-learning-based approaches, using two different datasets, namely the Caltech and TuSimple lane detection datasets, respectively. When tested against model-based approach, the proposed approach can offer as much as 5%, 12%, and 3% improvements on the true positive, false positive, and false positives per frame rates compared to the best alternative approach, respectively. When compared against a state-of-the-art machine learning technique, particularly against a supervised learning method, the proposed approach offers 57%, 31%, 4%, and 9&times, improvements on the false positive, false negative, accuracy, and frame rates. Furthemore, the proposed approach retains the explainability, or in other words, the cause of actions of the proposed approach can easily be understood or explained.

Obtaining accurate photometric redshift estimations is an important aspect of cosmology, remaining a prerequisite of many analyses. In creating novel methods to produce redshift estimations, there has been a shift towards using machine learning techniques. However, there has not been as much of a focus on how well different machine learning methods scale or perform with the ever-increasing amounts of data being produced. Here, we introduce a benchmark designed to analyse the performance and scalability of different supervised machine learning methods for photometric redshift estimation. Making use of the Sloan Digital Sky Survey (SDSS - DR12) dataset, we analysed a variety of the most used machine learning algorithms. By scaling the number of galaxies used to train and test the algorithms up to one million, we obtained several metrics demonstrating the algorithms' performance and scalability for this task. Furthermore, by introducing a new optimisation method, time-considered optimisation, we were able to demonstrate how a small concession of error can allow for a great improvement in efficiency. From the algorithms tested we found that the Random Forest performed best in terms of error with a mean squared error, MSE = 0.0042; however, as other algorithms such as Boosted Decision Trees and k-Nearest Neighbours performed incredibly similarly, we used our benchmarks to demonstrate how different algorithms could be superior in different scenarios. We believe benchmarks such as this will become even more vital with upcoming surveys, such as LSST, which will capture billions of galaxies requiring photometric redshifts., Comment: 9 pages, 6 figures, submitted to MNRAS

Structures of macromolecular assemblies derived from cryo-EM maps often contain errors that become more abundant with decreasing resolution. Despite efforts in the cryo-EM community to develop metrics for the map and atomistic model validation, thus far, no specific scoring metrics have been applied systematically to assess the interface between the assembly subunits. Here, we have assessed protein-protein interfaces in macromolecular assemblies derived by cryo-EM. To this end, we developed PI-score, a density-independent machine learning-based metric, trained using protein-protein interfaces’ features in high-resolution crystal structures. Using PI-score, we were able to identify errors at interfaces in the PDB-deposited cryo-EM structures (including SARS-CoV-2 complexes) and in the models submitted for cryo-EM targets in CASP13 and the EM model challenge. Some of the identified errors, especially at medium-to-low resolution structures, were not captured by density-based assessment scores. Our method can therefore provide a powerful complementary assessment tool for the increasing number of complexes solved by cryo-EM.

Structures of macromolecular assemblies derived from cryo-EM maps often contain errors that become more abundant with decreasing resolution. Despite efforts in the cryo-EM community to develop metrics for the map and atomistic model validation, thus far, no specific scoring metrics have been applied systematically to assess the interface between the assembly subunits. Here, we have assessed protein-protein interfaces in macromolecular assemblies derived by cryo-EM. To this end, we developed PI-score, a density-independent machine learning-based metric, trained using protein-protein interfaces’ features in high-resolution crystal structures. Using PI-score, we were able to identify errors at interfaces in the PDB-deposited cryo-EM structures (including SARS-CoV-2 complexes) and in the models submitted for cryo-EM targets in CASP13 and the EM model challenge. Some of the identified errors, especially at medium-to-low resolution structures, were not captured by density-based assessment scores. Our method can therefore provide a powerful complementary assessment tool for the increasing number of complexes solved by cryo-EM.

Classification of clutter, especially in the context of shore based radars, plays a crucial role in several applications. However, the task of distinguishing and classifying the sea clutter from land clutter has been historically performed using clutter models and/or coastal maps. In this paper, we propose two machine learning, particularly neural network, based approaches for sea-land clutter separation, namely the regularized randomized neural network (RRNN) and the kernel ridge regression neural network (KRR). We use a number of features, such as energy variation, discrete signal amplitude change frequency, autocorrelation performance, and other statistical characteristics of the respective clutter distributions, to improve the performance of the classification. Our evaluation based on a unique mixed dataset, which is comprised of partially synthetic clutter data for land and real clutter data from sea, offers improved classification accuracy. More specifically, the RRNN and KRR methods offer 98.50% and 98.75% accuracy, outperforming the conventional support vector machine and extreme learning based solutions.