Search papers, labs, and topics across Lattice.
This paper introduces an incremental evaluation and training paradigm for Relational Deep Learning (RDL) that addresses the limitations of static, single-episode benchmarks by modeling the continuous evolution of multi-tabular databases as temporal heterogeneous graphs. The authors demonstrate that traditional evaluation practices overlook the significant impact of temporal concept drifts on model performance, which can lead to suboptimal results in real-world applications. By implementing multiple incremental training regimes and a novel temporal evaluation metric, the study shows that fine-tuning RDL models with new data significantly enhances their adaptability and predictive accuracy compared to conventional training methods.
Incremental training can dramatically enhance the performance of RDL models, revealing that traditional evaluation methods miss critical shifts in data dynamics.
Relational Deep Learning (RDL) models multi-tabular databases as temporal heterogeneous graphs to enable end-to-end representation learning. However, prevailing RDL evaluation practices rely on static, single-episode dataset snapshots, overlooking the continuous, time-evolving nature of real-world databases. Consequently, current RDL benchmarks fail to capture how model performance changes as new data accumulates over time. To address this limitation, we introduce an incremental, multi-episode evaluation and training paradigm to assess and improve the temporal robustness and adaptability of state-of-the-art RDL models. Using established large-scale datasets, we examine data evolution and model training dynamics, demonstrating that temporal concept drifts occur in the majority of predictive tasks. We present multiple incremental training regimes for fine-tuning the models and demonstrate that transfer learning is both feasible and highly effective in the RDL setting. Alongside a new temporal evaluation metric that prioritizes near-future accuracy, we show that our incrementally fine-tuned models consistently outperform the standard, expensive, from-scratch trained baselines.
