arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1902.01780cs.LGcs.AI
收藏

通过通信递归学习决策树

Learning Decision Trees Recurrently Through Communication

University of Tübingen(图宾根大学);Tübingen AI Center(图宾根人工智能中心);MPI for Informatics(马克斯·普朗克信息学研究所);Saarland Informatics Campus(萨尔兰信息学园区);Wageningen University(瓦赫宁根大学);MPI for Intelligent Systems(马克斯·普朗克智能系统研究所);Max Planck Campus(马克斯·普朗克园区)

Stephan Alaniz, Diego Marcos, Bernt Schiele, Zeynep Akata

摘要:

在不牺牲决策算法的预测准确性的情况下,集成的可解释性具有极大地提高其对用户的价值的潜力。我们建议学习迭代二进制子决策,而不是直接为图像分配标签,从而在决策过程中引入稀疏性和透明度。我们的模型的关键方面是它能够建立一个决策树,其结构被编码到一个递归神经网络的记忆表示,由两个模型通过消息传递通信联合学习。此外,我们的模型以二进制属性的形式为每个决策分配语义含义,为用户提供简洁,语义和相关的合理化。在三个基准图像分类数据集上,包括大规模ImageNet,我们的模型生成了人类可解释的二进制决策序列,解释了网络的预测,同时保持了最先进的准确性。

英文摘要:

Integrated interpretability without sacrificing the prediction accuracy of decision making algorithms has the potential of greatly improving their value to the user. Instead of assigning a label to an image directly, we propose to learn iterative binary sub-decisions, inducing sparsity and transparency in the decision making process. The key aspect of our model is its ability to build a decision tree whose structure is encoded into the memory representation of a Recurrent Neural Network jointly learned by two models communicating through message passing. In addition, our model assigns a semantic meaning to each decision in the form of binary attributes, providing concise, semantic and relevant rationalizations to the user. On three benchmark image classification datasets, including the large-scale ImageNet, our model generates human interpretable binary decision sequences explaining the predictions of the network while maintaining state-of-the-art accuracy.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑