BLAST
收藏资源简介:
BLAST数据集是由中国科学院计算技术研究所人工智能安全国家重点实验室创建的,旨在通过平衡采样策略增强数据多样性。该数据集包含了从公开数据集中提取的3210亿个观测数据,并采用了一套全面的统计指标来描述时间序列模式。为了便于基于模式的采样,数据通过基于网格的划分进行隐式聚类。此外,通过集成网格采样和网格混合技术,BLAST确保了对多种模式的平衡和代表性的覆盖。实验结果表明,在BLAST上预训练的模型以更少的计算资源和训练标记实现了最先进的性能。
The BLAST dataset was developed by the State Key Laboratory of Artificial Intelligence Security, Institute of Computing Technology, Chinese Academy of Sciences, aiming to enhance data diversity through a balanced sampling strategy. It contains 321 billion observational data extracted from public datasets, and adopts a comprehensive set of statistical metrics to characterize time series patterns. To facilitate pattern-based sampling, the data is implicitly clustered via grid-based partitioning. Furthermore, by integrating grid sampling and grid mixing techniques, BLAST ensures balanced and representative coverage of diverse patterns. Experimental results demonstrate that models pre-trained on BLAST achieve state-of-the-art performance with fewer computational resources and training tokens.




