Learning feature interactions is crucial to success for large-scale CTR prediction in recommender systems and Ads ranking. Researchers and practitioners extensively proposed various neural network architectures for searching and modeling feature interactions. However, we observe that different datasets favor different neural network architectures and feature interaction types, suggesting that different feature interaction learning methods may have their own unique advantages. Inspired by this observation, we propose AdaEnsemble: a Sparsely-Gated Mixture-of-Experts (SparseMoE) architecture that can leverage the strengths of heterogeneous feature interaction experts and adaptively learns the routing to a sparse combination of experts for each example, allowing us to build a dynamic hierarchy of the feature interactions of different types and orders. To further improve the prediction accuracy and inference efficiency, we incorporate the dynamic early exiting mechanism for feature interaction depth selection. The AdaEnsemble can adaptively choose the feature interaction depth and find the corresponding SparseMoE stacking layer to exit and compute prediction from. Therefore, our proposed architecture inherits the advantages of the exponential combinations of sparsely gated experts within SparseMoE layers and further dynamically selects the optimal feature interaction depth without executing deeper layers. We implement the proposed AdaEnsemble and evaluate its performance on real-world datasets. Extensive experiment results demonstrate the efficiency and effectiveness of AdaEnsemble over state-of-the-art models.
翻译:研究人员和从业者广泛提出各种神经网络结构,供搜索和建模特征互动。然而,我们观察到,不同的数据集有利于不同的神经网络结构和特征互动类型,表明不同的特性互动学习方法可能具有独特的优势。我们根据这一观察,建议AdaEnsemble:一个可利用不同特性互动专家和适应性地学习专家优势的微小混合型(SparseMoE)结构,以利用不同特性互动专家和适应性化地学习专家的优势,将各种神经网络结构流到每个实例的分散组合,使我们能够建立不同类型和命令特征互动的动态等级。为了进一步提高预测准确性和推导效率,我们将动态早期退出机制纳入特征互动深度的选择。AdaEnsemble:可以适应性地选择特性互动深度,找到相应的SpressMoE堆叠叠层(SpareMoE)结构,以便从中找到相应的SprassimmoE层(Spregalmomomomentality)的精度组合,从而在Spreal-deal-deal-deal development abloa Ex层中进一步继承了我们提议的Spal-de-deal Ex层和Spreal-de Extravelopmental-demental 和Spalmentalmental-demental-demental-deal-demental-deal-deal-deal-deal-deal- a 和Spal-toal-toal-deal-deal-deal-deal-toalsalsal-toal 和Supal- a 和Spal-toal-deal- a 一级,我们提议的Slo 和Sppsal-to-to- a 一级,我们提议的Stoal- a 和S-toal-to-to- a 一级,在S-to-to-to-to-to-to-toal-toal-toal-toal-toal-toal-toal-toal-toal-toal-toal-to-to-to-to-to-toal-toal-to-to-toal- a a a a a a a a a a a a a a a a a a a a a a