MetaCC
收藏资源简介:
MetaCC是由三星人工智能中心剑桥分部和德克萨斯大学奥斯汀分校的研究团队开发的一个用于元学习的信道编码基准数据集。该数据集包含五个信道模型家族,涵盖了从合成到实际软件定义无线电的数据,旨在研究元学习在面对任务分布广度和转移时的性能。MetaCC数据集通过精细控制和测量任务分布的复杂性和转移,为研究社区提供了一个工具,以探讨元学习的能力和局限性,并推动研究实用的鲁棒和有效的元学习者。
MetaCC is a channel coding benchmark dataset for meta-learning, developed by research teams from Samsung AI Center Cambridge and The University of Texas at Austin. This dataset encompasses five channel model families, covering data spanning from synthetic settings to real-world software-defined radio measurements. It is designed to investigate the performance of meta-learning algorithms in the face of task distribution breadth and distribution shift. The MetaCC dataset enables fine-grained control and measurement of task distribution complexity and distribution shift, providing the research community with a valuable tool to explore the capabilities and limitations of meta-learning, and to advance the development of practical, robust and efficient meta-learners.




