图像情境下的数字序列逻辑学习  

Number sequence logic learning in image context

作　　者：梁慧 曹峰 钱宇华[1,2,3] 郭倩 梁新彦[1,3] LIANG Hui;CAO Feng;QIAN Yuhua;GUO Qian;LIANG Xinyan(Research Institute of Big Data Science and Industry,Shanxi University,Taiyuan 030006,China;Key Laboratory of Computational Intelligence and Chinese Information Processing of Ministry of Education,Shanxi University,Taiyuan 030006,China;School of Computer and Information Technology,Shanxi University,Taiyuan 030006,China)

机构地区：[1]山西大学大数据科学与产业研究院,山西太原030006 [2]山西大学计算智能与中文信息处理教育部重点实验室,山西太原030006 [3]山西大学计算机与信息技术学院,山西太原030006

出　　处：《智能系统学报》2019年第6期1189-1198,共10页CAAI Transactions on Intelligent Systems

基　　金：国家自然科学基金项目(61672332,61432011,U1435212,61872226);山西省海外归国人员研究项目(2017023);山西省自然科学基金计划资助项目(201701D121052)

摘　　要：针对未知的数字和规则的模式构建问题,本文提供了一种从图像角度解决数字序列逻辑学习问题的手段。该方法是在计算机不知道图像间关系和图像内包含的内容的意义的前提下,让计算机自主地学习出其中包含的内在逻辑模式,从而进行数字序列的预测。本文构建了4个大型数据集:Linear序列、Multiplication序列、Fio序列和Nested序列,然后使用几种代表性的深度神经网络来完成数字序列逻辑学习任务,并对实验结果加以分析比较,事实证明,本文所提出的方法在一定程度上可以解决未知的数字和规则的模式构建问题,这为一系列未知逻辑模式构建任务提供了一种可能性。To solve the problem of pattern construction of unknown numbers and rules,in this paper,we provide a meth-od to solve the problems of number sequence logic learning from the image perspective.The method allows the com-puter to automatically learn the inherent logic pattern without prior knowledge of the meaning of the image content or of the relationship between images so as to predict the number sequence.Four large datasets were constructed:linear se-quences,multiplication sequences,fio sequences,and nested sequences,and then several representative deep neural net-works were used to complete the number sequence logic learning task.By analyzing the experimental results,the meth-od was found capable of solving the problem of pattern construction for unknown numbers and rules to a certain extent,which will provide a potential solution for a series of unknown logic pattern construction tasks.

关 键 词：人工智能 逻辑推理 逻辑学习 深度学习 数字序列 图像处理 神经网络 模式构建 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]