Feature Selection for Facebook Feed Ranking System via a Group-Sparsity-Regularized Training Algorithm
Feature Selection for Facebook Feed Ranking System via a Group-Sparsity-Regularized Training Algorithm
复制标题
DOI：
10.1145/3357384.3358114
复制
发表时间：
2019-11
期刊：
影响因子：
--
通讯作者：
Xiuyan Ni;Yang Yu;Pengxiang Wu;Youlin Li;Shaoliang Nie;Qichao Que;Chao Chen
中科院分区：
文献类型：
--
作者：
Xiuyan Ni;Yang Yu;Pengxiang Wu;Youlin Li;Shaoliang Nie;Qichao Que;Chao Chen
在现代生产平台中，大规模在线学习模型被应用于非常高维的数据。为了节省计算资源，重要的是要有一个有效的算法来选择最重要的功能，从一个巨大的特征池。在本文中，我们提出了一种新的神经网络适合的特征选择算法，在训练过程中从输入层中选择重要的特征。而不是直接正则化的训练损失，我们注入组稀疏正则化的（随机）训练算法。特别地，我们引入了一个组稀疏范数的近似正则化随机梯度下降算法。为了充分评估实际性能，我们将我们的方法应用于Facebook News Feed数据集，并与使用传统正则化器的最新技术相比取得了良好的性能。
In modern production platforms, large scale online learning models are applied to data of very high dimension. To save computational resource, it is important to have an efficient algorithm to select the most significant features from an enormous feature pool. In this paper, we propose a novel neural-network-suitable feature selection algorithm, which selects important features from the input layer during training. Instead of directly regularizing the training loss, we inject group-sparsity regularization into the (stochastic) training algorithm. In particular, we introduce a group sparsity norm into the proximally regularized stochastical gradient descent algorithm. To fully evaluate the practical performance, we apply our method to Facebook News Feed dataset, and achieve favorable performance compared with state-of-the-arts using traditional regularizers.