yasoz / synmob Goto Github PK

Hi, thank you for providing the synthetic data generated! For the four attributes [trip distance, trip time, departure time, sample points] (I assume these are descriptions of the synthetic trajectories generated so it should be different from the attributes of original trajectories that are used to generate synthetic ones), I understand that trip distance, trip time, sample points can be calculated from the synthetic trajectory by calculating distance between points and counting the number of points in a trajectory. But I was wondering how departure time is calculated. I was also wondering why trip total time is length of points *3, Thank you!

您好！首先祝贺您首次提出了使用Diffusion生成时空轨迹的方法，在拜读了您的论文以及下载了您的数据集参考后，我产生了几个问题想向您请教一下，恳请得到您的回答！
问：

1. 我下载了其中一个西安城市的轨迹数据集，发现将所有轨迹可视化出来后，第1183条轨迹存在着点与点之间经纬度跨度大、且各点呈现不规律的散点分布，请问是原本数据集中本就存在无序的点，还是说diffusion在生成过程中的缺陷导致的？
2. 请问生成的数据集和原始数据集属于同一个分布的情况下，生成的轨迹具体数值是否和原始轨迹具体数据一致？
3. 请问作者在设计噪声预测网络时为什么选用了ResNet的设计，而不使用U-Net作为噪声预测网络呢？