Touch the Core: Exploring Task Dependence Among Hybrid Targets for Recommendation

TL;DR

This paper tackles hybrid-target learning in recommender systems by modeling explicit and implicit dependence between a core continuous target and preceding discrete actions. It proposes HTLNet, featuring a Label Embedding Unit (LEU) to transfer label information and an Information Fusion Unit (IFU) to adaptively fuse prior task signals, together with a gradient-adaptation optimization to stabilize training. Empirical results on two public datasets and one industrial dataset show HTLNet achieving state-of-the-art performance across core and auxiliary tasks, with online A/B tests confirming gains in purchase-related metrics. The work highlights the importance of designing task-dependence priors and a tailored optimization strategy for robust hybrid-target learning in large-scale recommender systems.

Abstract

As user behaviors become complicated on business platforms, online recommendations focus more on how to touch the core conversions, which are highly related to the interests of platforms. These core conversions are usually continuous targets, such as \textit{watch time}, \textit{revenue}, and so on, whose predictions can be enhanced by previous discrete conversion actions. Therefore, multi-task learning (MTL) can be adopted as the paradigm to learn these hybrid targets. However, existing works mainly emphasize investigating the sequential dependence among discrete conversion actions, which neglects the complexity of dependence between discrete conversions and the final continuous conversion. Moreover, simultaneously optimizing hybrid tasks with stronger task dependence will suffer from volatile issues where the core regression task might have a larger influence on other tasks. In this paper, we study the MTL problem with hybrid targets for the first time and propose the model named Hybrid Targets Learning Network (HTLNet) to explore task dependence and enhance optimization. Specifically, we introduce label embedding for each task to explicitly transfer the label information among these tasks, which can effectively explore logical task dependence. We also further design the gradient adjustment regime between the final regression task and other classification tasks to enhance the optimization. Extensive experiments on two offline public datasets and one real-world industrial dataset are conducted to validate the effectiveness of HTLNet. Moreover, online A/B tests on the financial recommender system also show that our model has improved significantly. Our implementation is available here\footnote{\url{https://github.com/fuyuanlyu/HTLNet}}.

Figures (8)

• Figure 1: Illustration of hybrid target learning paradigm in various business platforms. The red one denotes core targets.
• Figure 2: The overall framework of HTLNet. LEU represents the Latent Embedding Unit, which encodes label information, and the Information Fusion Unit, which adaptive infuses all preceding tasks.
• Figure 3: Illustration of Label Embedding Unit and Information Fusion Unit. $\otimes$ denotes element-wise multiplication. Note that $p^t_0+p^t_1=1$.
• Figure 4: Comparison of the different operations in IFU.
• Figure 5: The illustrative case of the effect of optimization on the KuaiRand dataset. The legend indicates the different tasks. The Y-axis is the average shared embedding gradient magnitude over all mini-batch iterations in one training step.