S^(2)ANet:Combining local spectral and spatial point grouping for point cloud processing

Background Despite the recent progress in 3D point cloud processing using deep convolutional neural networks,the inability to extract local features remains a challenging problem.In addition,existing methods consider only the spatial domain in the feature extraction process.Methods In this paper,we propose a spectral and spatial aggregation convolutional network(S^(2)ANet),which combines spectral and spatial features for point cloud processing.First,we calculate the local frequency of the point cloud in the spectral domain.Then,we use the local frequency to group points and provide a spectral aggregation convolution module to extract the features of the points grouped by the local frequency.We simultaneously extract the local features in the spatial domain to supplement the final features.Results S^(2)ANet was applied in several point cloud analysis tasks;it achieved stateof-the-art classification accuracies of 93.8%,88.0%,and 83.1%on the ModelNet40,ShapeNetCore,and ScanObjectNN datasets,respectively.For indoor scene segmentation,training and testing were performed on the S3DIS dataset,and the mean intersection over union was 62.4%.Conclusions The proposed S^(2)ANet can effectively capture the local geometric information of point clouds,thereby improving accuracy on various tasks.

Accelerated strategy for fast ion conductor materials screening and optimal doping scheme exploration

Fast ion conductor materials screening based on high-throughput calculations involves enormous computing tasks.The process usually includes structural optimization,energy calculation,charge analysis and ionic migration performance estimation.The first one involves looking for the equilibrium atomic positions in huge amount of candidate compounds or derivative structures,and the computational cost is always high because of the task-intensive features.The last one relates to the kinetic problems,for which the time-consuming transition state theory and the molecular dynamics are the main simulation methods.In this work,two predictive models,ionic migration activation energy model and structural optimization model,are developed based on machine learning(ML)techniques to accelerate the process of estimating activation energy and relaxing the doped crystal structures,respectively.By training 3136 energy barrier data calculated by bond valence(BV)method,an ionic migration activation energy model(Ea model)with mean absolute error(MAE)of 0.26 eV on testing data set is obtained.We apply this model and filter LiBiOS as a promising fast Li^(+)conductor from 49 Licontaining hetero-anionic compounds.Although the model-predicted result shows relatively low energy barrier,further analysis indicates that the high carrier formation energy restricts the ionic transportability.Therefore,we substitute fractional Li^(+)with Mg^(2+)in LiBiOS to relieve the large difficulty of forming carriers in the structure.In order to fast explore the optimal doping scheme,we develop the structural optimization model(E-f model)containing the ML-based energy and force prediction to accelerate the structural optimization under various LieMg ratio and doping configurations.Decent doping scheme Li_(1-2x)Mg_(x)BiOS(x=0.1875)shows much better Li^(+)migration performance compared with LiBiOS without substitution.This method of screening fast ion conductor materials and finding optimal doping scheme will extremely accelerate materials explorations.