The capability of making interpretable and self-explanatory decisions is essential for developing responsible machine learning systems. In this work, we study the learning to explain problem in the scope of inductive logic programming (ILP). We propose Neural Logic Inductive Learning (NLIL), an efficient differentiable ILP framework that learns first-order logic rules that can explain the patterns in the data. In experiments, compared with the state-of-the-art methods, we find NLIL can search for rules that are x10 times longer while remaining x3 times faster. We also show that NLIL can scale to large image datasets, i.e. Visual Genome, with 1M entities.
翻译:做出可解释和自解决定的能力对于发展负责任的机器学习系统至关重要。 在这项工作中,我们研究学习如何解释感化逻辑程序(ILP)范围的问题。我们提出神经逻辑感化学习(NLIL),这是一个高效的、可区分的ILP框架,可以学习能够解释数据模式的第一阶逻辑规则。在实验中,与最先进的方法相比,我们发现NLIL可以寻找比X10倍长而剩下的x3倍更快的规则。我们还表明,NLIL可以向大型图像数据集(即视觉基因组)扩展,与1M实体相比。