World Population Growth
收藏www.kaggle.com2023-03-16 更新2025-01-09 收录
下载链接：
https://www.kaggle.com/cici118/world-population-growth
下载链接
链接失效反馈官方服务：
资源简介：
# Featured at "Prior knowledge meets Neural ODEs: a two-stage training method for improved explainability" a Tiny Paper @ ICLR 2023 https://openreview.net/forum?id=p7sHcNt_tqo
The World Population Growth Dataset was synthesized and has one feature, world population. The
goal is to predict the population at each point in time.
There are three tasks available:
-reconstruction: evaluate the performance at predicting the same time steps used for training;
-extrapolation: evaluate the performance at predicting for a longer time horizon than the one used for training;
-completion: evaluate the performance at predicting time steps in between the ones used for training.
The dataset was created by solving the following ODE:
收录于《先验知识邂逅神经常微分方程：一种提升可解释性的两阶段训练方法》的Tiny Paper，发表于2023年ICLR会议。该数据集系由世界人口增长数据集合成，包含单一特征——世界人口。研究目标在于预测每一个时间点的人口数量。该数据集涵盖了三项任务：
- 重构：评估在预测与训练所用时间步长相同的性能；
- 外推：评估在预测训练所用时间范围之外的更长时间跨度的性能；
- 补全：评估在预测训练所用时间步长之间的性能。数据集的构建基于以下常微分方程（ODE）的求解：[常微分方程图](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F2406836%2F51dbf75dd99dfb0856183cea296409d2%2FScreenshot%20from%202023-03-22%2014-48-34.png?generation=1679496528936237&alt=media)
提供机构：
Kaggle
