Search

Robust and scalable function evaluation at any arbitrary point in the finite/spectral element mesh is required for querying the partial differential equation solution at points of interest, comparison of solution between different meshes, and Lagrangian particle tracking. This is a challenging problem, particularly for high-order unstructured meshes partitioned in parallel with MPI, as it requires identifying the element that overlaps a given point and computing the corresponding reference space coordinates. We present a robust and efficient technique for general field evaluation in large-scale high-order meshes with quadrilaterals and hexahedra. In the proposed method, a combination of globally partitioned and processor-local maps are used to first determine a list of candidate MPI ranks, and then locally candidate elements that could contain a given point. Next, element-wise bounding boxes further reduce the list of candidate elements. Finally, Newton's method with trust region is used to determine the overlapping element and corresponding reference space coordinates. Since GPU-based architectures have become popular for accelerating computational analyses using meshes with tensor-product elements, specialized kernels have been developed to utilize the proposed methodology on GPUs. The method is also extended to enable general field evaluation on surface meshes. The paper concludes by demonstrating the use of proposed method in various applications ranging from mesh-to-mesh transfer during r-adaptivity to Lagrangian particle tracking., Comment: 52 pages, 17 figures, 1 table

Secondary fragmentation of an impulsively accelerated drop depends on fluid properties and velocity of the ambient. The critical Weber number $(\mathit{We}_\mathit{cr})$, the minimum Weber number at which a drop undergoes non-vibrational breakup, depends on fluid density ratio $(\rho)$, the drop $(\mathit{Oh}_d)$, and the ambient $(\mathit{Oh}_o)$ Ohnesorge numbers. The current study uses VoF based interface-tracking multiphase flow simulations to quantify the effect of different non-dimensional groups on the threshold at which secondary fragmentation occur. For $\mathit{Oh}_d \leq 0.1$, a decrease in $\mathit{Oh}_d$ significantly influences the breakup morphology, plume formation, and the resulting $\mathit{We}_\mathit{cr}$. $\rho$ and $\mathit{Oh}_o$, were found to influence the balance between the pressure differences between the poles and the periphery, and the shear stresses on the upstream surface. These external forces induce flow inside the initially spherical drop, resulting in deformation into pancakes and eventually the breakup morphology of forward/backward bag. The evolution pathways of the drop morphology based on their non-dimensional groups have been charted. With the inclusion of the new observations, the traditional $\mathit{We}_\mathit{cr}-\mathit{Oh}_d$ plot, used for illustrating the dependence of critical Weber number on $\mathit{Oh}_d$, was found to be inadequate in predicting the minimum initial $\mathit{We}$ required to undergo fragmentation. A new non-dimensional parameter $C_{breakup}$ is derived based on the competition between forces driving drop deformation and the forces resisting drop deformation. Tested on the available experimental data and current simulations, $C_{breakup}$ is found to be a robust predictor for the threshold of drop fragmentation., Comment: 40 pages, 22 figures

There has a been a significant rise in the use of Community Question Answering sites (CQAs) over the last decade owing primarily to their ability to leverage the wisdom of the crowd. Duplicate questions have a crippling effect on the quality of these sites. Tackling duplicate questions is therefore an important step towards improving quality of CQAs. In this regard, we propose two neural network based architectures for duplicate question detection on Stack Overflow. We also propose explicitly modeling the code present in questions to achieve results that surpass the state of the art.

Datacenter-scale AI training clusters consisting of thousands of domain-specific accelerators (DSA) are used to train increasingly-complex deep learning models. These clusters rely on a data storage and ingestion (DSI) pipeline, responsible for storing exabytes of training data and serving it at tens of terabytes per second. As DSAs continue to push training efficiency and throughput, the DSI pipeline is becoming the dominating factor that constrains the overall training performance and capacity. Innovations that improve the efficiency and performance of DSI systems and hardware are urgent, demanding a deep understanding of DSI characteristics and infrastructure at scale. This paper presents Meta's end-to-end DSI pipeline, composed of a central data warehouse built on distributed storage and a Data PreProcessing Service that scales to eliminate data stalls. We characterize how hundreds of models are collaboratively trained across geo-distributed datacenters via diverse and continuous training jobs. These training jobs read and heavily filter massive and evolving datasets, resulting in popular features and samples used across training jobs. We measure the intense network, memory, and compute resources required by each training job to preprocess samples during training. Finally, we synthesize key takeaways based on our production infrastructure characterization. These include identifying hardware bottlenecks, discussing opportunities for heterogeneous DSI hardware, motivating research in datacenter scheduling and benchmark datasets, and assimilating lessons learned in optimizing DSI infrastructure., Comment: In The 49th Annual International Symposium on Computer Architecture (ISCA 2022)

Pre-trained word embeddings are the primary method for transfer learning in several Natural Language Processing (NLP) tasks. Recent works have focused on using unsupervised techniques such as language modeling to obtain these embeddings. In contrast, this work focuses on extracting representations from multiple pre-trained supervised models, which enriches word embeddings with task and domain specific knowledge. Experiments performed in cross-task, cross-domain and cross-lingual settings indicate that such supervised embeddings are helpful, especially in the low-resource setting, but the extent of gains is dependent on the nature of the task and domain. We make our code publicly available., Comment: Appeared in 2nd Learning from Limited Labeled Data (LLD) Workshop at ICLR 2019

We sequenced genomes from a $\sim$7,000 year old early farmer from Stuttgart in Germany, an $\sim$8,000 year old hunter-gatherer from Luxembourg, and seven $\sim$8,000 year old hunter-gatherers from southern Sweden. We analyzed these data together with other ancient genomes and 2,345 contemporary humans to show that the great majority of present-day Europeans derive from at least three highly differentiated populations: West European Hunter-Gatherers (WHG), who contributed ancestry to all Europeans but not to Near Easterners; Ancient North Eurasians (ANE), who were most closely related to Upper Paleolithic Siberians and contributed to both Europeans and Near Easterners; and Early European Farmers (EEF), who were mainly of Near Eastern origin but also harbored WHG-related ancestry. We model these populations' deep relationships and show that EEF had $\sim$44% ancestry from a "Basal Eurasian" lineage that split prior to the diversification of all other non-African lineages.