For effective flow visualization, identifying representative lines or surfaces is an important problem which has been studied. However, no work can solve the for both and surfaces. In this paper, we present FlowNet, a single deep learning framework clustering selection of streamlines stream Given collection generated from field data set, our approach converts them into binary volumes then employs autoencoder to learn their respective latent feature descriptors. These descriptors are used...

We present TSR-TVD, a novel deep learning framework that generates temporal super-resolution (TSR) of time-varying data (TVD) using adversarial learning. TSR-TVD is the first work applies recurrent generative network (RGN), combination neural (RNN) and (GAN), to generate high-resolution volume sequences from low-resolution ones. The design includes generator discriminator. takes pair volumes as input outputs synthesized intermediate sequence through forward backward predictions....

Deep learning has been applied successfully to many biomedical image segmentation tasks. However, due the diversity and complexity of data, manual annotation for training common deep models is very timeconsuming labor-intensive, especially because normally only experts can annotate data well. Human are often involved in a long iterative process annotation, as active type schemes. In this paper, we propose representative (RA), new framework reducing effort segmentation. RA uses unsupervised...

Since 2016, we have witnessed the tremendous growth of artificial intelligence+visualization (AI+VIS) research. However, existing survey articles on AI+VIS focus visual analytics and information visualization, not scientific visualization (SciVis). In this article, related deep learning (DL) works in SciVis, specifically direction DL4SciVis: designing DL solutions for solving SciVis problems. To stay focused, primarily consider that handle scalar vector field data but exclude mesh data. We...

We present SSR-VFD, a novel deep learning framework that produces coherent spatial super-resolution (SSR) of three-dimensional vector field data (VFD). SSR-VFD is the first work advocates machine approach to generate high-resolution fields from low-resolution ones. The core lies in use three separate neural nets take components as input and jointly output synthesized field. To capture coherence, we into account magnitude angle losses network optimization. Our method can situ scenario where...

Indoleamine 2,3-dioxygenase (IDO), as a crucial immunomodulatory target, can play pivotal role in regulating immune function. In this study, the ligand-based pharmacophore model (HDAA) and receptor-based (HHDA) were respectively constructed, which could comprehensively reflect spatial distribution characteristics of core groups IDO inhibitors. Subsequently, these models utilized to guide design novel inhibitor through skeleton combination pathway. The corresponding molecular structures...

Abstract The prevention and management of coal mine roof accidents remain challenging issues because it is difficult to evaluate quantify the interaction effects disaster hazard factors objectively. This paper proposes a novel approach: combining information entropy surrogate model—and applies Sobol’s method, aiming solve obtain factors’ 1th global sensitivity value without human intervention. results show that: (1) complex logical relationships interactions can be transformed into...

We present SSR-TVD, a novel deep learning framework that produces coherent spatial super-resolution (SSR) of time-varying data (TVD) using adversarial learning. In scientific visualization, SSR-TVD is the first work applies generative network (GAN) to generate high-resolution volumes for three-dimensional sets. The design includes generator and two discriminators (spatial temporal discriminators). takes low-resolution volume as input outputs synthesized volume. To capture coherence in...

We present STNet, an end-to-end generative framework that synthesizes spatiotemporal super-resolution volumes with high fidelity for time-varying data. STNet includes two modules: a generator and discriminator. The input to the is low-resolution at both ends, output intermediate two-ending volumes. discriminator, leveraging convolutional long short-term memory, accepts sequence as predicts conditional score each volume based on its spatial (the itself) temporal previous volumes) information....

Cloud computing affects the development of mobile learning with its unique advantages. This article first sums up traditional mode learning, analyses characteristics various patterns, sum cloud computing, and then make under a environment.

We present a new approach for streamline-based flow field representation and reduction. Our method can work in the situ visualization setting by tracing streamlines from each time step of simulation storing compressed post hoc analysis where users afford longer reconstruction higher quality using decompressed streamlines. At heart our is deep learning vector that takes traced original fields as input applies two-stage process to reconstruct high-quality fields. To demonstrate effectiveness...

We present V2V, a novel deep learning framework, as general-purpose solution to the variable-to-variable (V2V) selection and translation problem for multivariate time-varying data (MTVD) analysis visualization. V2V leverages representation algorithm identify transferable variables utilizes Kullback-Leibler divergence determine source target variables. It then uses generative adversarial network (GAN) learn mapping from variable via adversarial, volumetric, feature losses. takes pairs of time...

Although deep learning has demonstrated its capability in solving diverse scientific visualization problems, it still lacks generalization power across different tasks. To address this challenge, we propose CoordNet, a single coordinate-based framework that tackles various tasks relevant to time-varying volumetric data without modifying the network architecture. The core idea of our approach is decompose task inputs and outputs into unified representation (i.e., coordinates values) learn...

We propose VFR-UFD, a new deep learning framework that performs vector field reconstruction (VFR) for unsteady flow data (UFD). Given integral lines (i.e., streamlines), we first generate low-quality UFD via diffusion. VFR-UFD then leverages convolutional neural network to reconstruct spatiotemporally coherent, high-quality UFD. The core of lies in recurrent residual blocks iteratively refine and denoise the input fields at different scales, both locally globally. take consecutive time steps...

Medicinal plants are the primary sources for discovery of novel medicines and basis ethnopharmacological research. While existing studies mainly focus on chemical compounds, there is little research about functions other contents in medicinal plants. Extracellular vesicles (EVs) functionally active, nanoscale, membrane-bound secreted by almost all eukaryotic cells. Intriguingly, plant-derived extracellular (PDEVs) also have been implicated to play an important role therapeutic application....

The usage of deep generative models for image compression has led to impressive performance gains over classical codecs while neural video is still in its infancy. Here, we propose an end-to-end, modeling approach compress temporal sequences with a focus on video. Our builds upon variational autoencoder (VAE) sequential data and combines them recent work compression. jointly learns transform the original sequence into lower-dimensional representation as well discretize entropy code this...

Stein variational gradient decent (SVGD) has been shown to be a powerful approximate inference algorithm for complex distributions. However, the standard SVGD requires calculating of target density and cannot applied when is unavailable. In this work, we develop gradient-free variant (GF-SVGD), which replaces true with surrogate gradient, corrects induced bias by re-weighting gradients in proper form. We show that our GF-SVGD can viewed as special choice kernel, hence directly inherits...