ReCODE: Modeling Repeat Consumption with Neural ODE

TL;DR

ReCODE addresses the challenge of repeat consumption in recommender systems by eschewing predefined time-gap distributions and instead modeling dynamics with neural ODEs. It decomposes the recommendation score into a static preference component $\hat{R}_{u,i}^\text{sta}$ and a dynamic repeat-aware component $\hat{R}_{u,i}^\text{rep}$, yielding $\hat{R}_{u,i}=\hat{R}_{u,i}^\text{sta}+\hat{R}_{u,i}^\text{rep}$, where the dynamic part evolves from an encoder-generated initial state through a neural ODE driven by historical time gaps. The framework is model-agnostic and can retrofit various backbones (MF, NCF, GRU4Rec, SASRec, etc.), with a pairwise ranking loss guiding training. Empirical results on two real-world music datasets show consistent improvements across base models and higher gains at larger repeat ratios, highlighting the practical impact of capturing repeat consumption patterns in live systems.

Abstract

In real-world recommender systems, such as in the music domain, repeat consumption is a common phenomenon where users frequently listen to a small set of preferred songs or artists repeatedly. The key point of modeling repeat consumption is capturing the temporal patterns between a user's repeated consumption of the items. Existing studies often rely on heuristic assumptions, such as assuming an exponential distribution for the temporal gaps. However, due to the high complexity of real-world recommender systems, these pre-defined distributions may fail to capture the intricate dynamic user consumption patterns, leading to sub-optimal performance. Drawing inspiration from the flexibility of neural ordinary differential equations (ODE) in capturing the dynamics of complex systems, we propose ReCODE, a novel model-agnostic framework that utilizes neural ODE to model repeat consumption. ReCODE comprises two essential components: a user's static preference prediction module and the modeling of user dynamic repeat intention. By considering both immediate choices and historical consumption patterns, ReCODE offers comprehensive modeling of user preferences in the target context. Moreover, ReCODE seamlessly integrates with various existing recommendation models, including collaborative-based and sequential-based models, making it easily applicable in different scenarios. Experimental results on two real-world datasets consistently demonstrate that ReCODE significantly improves the performance of base models and outperforms other baseline methods.

Figures (3)

• Figure 1: Repeat consumption time gap distributions of different user-item pairs on MMTD dataset.
• Figure 2: Architecture of our proposed ReCODE framework.
• Figure 3: Performance comparisons (Recall@50) w.r.t. different repeat ratios on MMTD and Nowplaying-RS datasets. The shaded area represents the 95% confidence intervals of $t$-distribution obtained from the five experimental runs.