arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2002.04322cs.LGstat.ML
收藏

基于神经网络的特征交互学习局部最优问题研究

A study of local optima for learning feature interactions using neural networks

Mathematics Department, Florida State University(佛罗里达州立大学数学系);Statistics Department, Florida State University(佛罗里达州立大学统计系)

Yangzi Guo, Adrian Barbu

摘要:

在生物信息学、高能物理、配电等诸多领域,希望学习非线性模型,其中选择少量变量并且明确地对它们之间的相互作用建模以预测响应。原则上,神经网络(NN)可以完成这项任务,因为它们可以很好地模拟非线性特征交互。然而,神经网络需要大量的训练数据才能有良好的泛化能力。在本文中,我们研究的datastarved制度,一个神经网络的训练数据相对较少的训练。为此,我们研究了神经网络的特征选择,这是已知的,以提高线性模型的泛化。作为具有特征选择和特征交互的数据的极端情况,我们研究了具有不相关变量的XOR类数据。我们通过实验观察到异或类数据上的交叉熵损失函数存在许多不等价的局部最优解,并且局部最优解的数量随着无关变量的数量呈指数增长。为了处理局部极小值和特征选择,我们提出了一个节点修剪和特征选择算法,提高了神经网络的能力,找到更好的局部极小值,即使有不相关的变量。最后,我们证明了神经网络在真实数据集上的性能可以通过修剪来提高,在少量特征上获得紧凑的网络,具有良好的预测性和可解释性。

英文摘要:

In many fields such as bioinformatics, high energy physics, power distribution, etc., it is desirable to learn non-linear models where a small number of variables are selected and the interaction between them is explicitly modeled to predict the response. In principle, neural networks (NNs) could accomplish this task since they can model non-linear feature interactions very well. However, NNs require large amounts of training data to have a good generalization. In this paper we study the datastarved regime where a NN is trained on a relatively small amount of training data. For that purpose we study feature selection for NNs, which is known to improve generalization for linear models. As an extreme case of data with feature selection and feature interactions we study the XOR-like data with irrelevant variables. We experimentally observed that the cross-entropy loss function on XOR-like data has many non-equivalent local optima, and the number of local optima grows exponentially with the number of irrelevant variables. To deal with the local minima and for feature selection we propose a node pruning and feature selection algorithm that improves the capability of NNs to find better local minima even when there are irrelevant variables. Finally, we show that the performance of a NN on real datasets can be improved using pruning, obtaining compact networks on a small number of features, with good prediction and interpretability.

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。

↑