3D Charged Particles Dataset - Roto-translated Local Coordinate Frames for Interacting Dynamical Systems

This repository contains the "<strong>3D charged particles</strong>" dataset from the paper <strong>Roto-translated Local Coordinate Frames for Interacting Dynamical Systems</strong><br> Miltiadis Kofinas, Naveen Shankar Nagaraja, Efstratios Gavves<br> NeurIPS 2021<br> https://arxiv.org/abs/2110.14961<br> https://github.com/mkofinas/locs It contains simulations of trajectories of 5 charged particles in 3 dimensions, interacting via Coulomb forces. There are 30,000 simulations for training, 5,000 for validation, and 5,000 for testing. Train and validation simulations last for 99 timesteps, while test simulations last for 99 timesteps. The features comprise positions and velocities of particles, while edges describe the product of pairwise charges.本仓库收录来自论文《旋移局部坐标系用于交互动力学系统（Roto-translated Local Coordinate Frames for Interacting Dynamical Systems）》的3D带电粒子（3D charged particles）数据集，作者为Miltiadis Kofinas、Naveen Shankar Nagaraja、Efstratios Gavves，发表于2021年NeurIPS会议。相关资源链接：https://arxiv.org/abs/2110.14961、https://github.com/mkofinas/locs。 该数据集包含5个带电粒子在三维空间中通过库仑力相互作用的轨迹模拟样本：训练集共30000组，验证集与测试集各含5000组。训练与验证模拟样本的时长为99个时间步，测试模拟样本同样包含99个时间步。数据集特征涵盖粒子的位置与速度信息，边特征用于描述粒子两两间的电荷乘积。