中文

用多层关系网络求解瑞文推理测验

机器学习 2020-03-27 v1 机器学习

摘要

瑞文推理测验(Raven's Progressive Matrices)是一项最初设计用于测试人类认知能力的基准。近来它已被改编用于测试机器学习系统的关系推理能力。为此建立了所谓的程序生成矩阵(Procedurally Generated Matrices)数据集,该数据集迄今仍是最困难的关系推理基准之一。本文中我们表明深度神经网络能够求解该基准,通过将 Wild Relation Networks 与多层关系网络(Multi-Layer Relation Networks)相结合并引入幅度编码(Magnitude Encoding,一种为后期融合架构设计的编码方案),在先前 62.6% 的最优性能(state-of-the-art)基础上达到了 98.0% 的准确率。

关键词

引用

@article{arxiv.2003.11608,
  title  = {Solving Raven's Progressive Matrices with Multi-Layer Relation Networks},
  author = {Marius Jahrens and Thomas Martinetz},
  journal= {arXiv preprint arXiv:2003.11608},
  year   = {2020}
}

备注

6 pages, 6 figures, to be published in the Proceedings of the IJCNN 2020, source code available at http://webmail.inb.uni-luebeck.de/exchange-supplement/PGM_MLRN_supplementary.zip