Robust Deep Hawkes Process under Label Noise of Both Event and Occurrence

TL;DR

This work tackles robustness of deep Hawkes processes under label noise affecting both event types and occurrence times. RDHP combines a Generalized Cross Entropy loss for event-type prediction, a sparse overparameterization mechanism for time-prediction, and a loss-reweighting framework to mitigate compounded labeling errors in intensity learning, validated on synthetic benchmarks and real OSAHS data. Results show RDHP outperforms SAHP and THP in both event-type classification (macro $F_1$) and event-time regression (RMSE) across noisy conditions, demonstrating strong practical relevance for medical time-series analysis. The approach offers a principled path to deploy deep Hawkes models in real-world, noisy clinical settings by explicitly addressing both sources and interactions of label noise, with potential applicability to a broader class of point-process models.

Abstract

Integrating deep neural networks with the Hawkes process has significantly improved predictive capabilities in finance, health informatics, and information technology. Nevertheless, these models often face challenges in real-world settings, particularly due to substantial label noise. This issue is of significant concern in the medical field, where label noise can arise from delayed updates in electronic medical records or misdiagnoses, leading to increased prediction risks. Our research indicates that deep Hawkes process models exhibit reduced robustness when dealing with label noise, particularly when it affects both event types and timing. To address these challenges, we first investigate the influence of label noise in approximated intensity functions and present a novel framework, the Robust Deep Hawkes Process (RDHP), to overcome the impact of label noise on the intensity function of Hawkes models, considering both the events and their occurrences. We tested RDHP using multiple open-source benchmarks with synthetic noise and conducted a case study on obstructive sleep apnea-hypopnea syndrome (OSAHS) in a real-world setting with inherent label noise. The results demonstrate that RDHP can effectively perform classification and regression tasks, even in the presence of noise related to events and their timing. To the best of our knowledge, this is the first study to successfully address both event and time label noise in deep Hawkes process models, offering a promising solution for medical applications, specifically in diagnosing OSAHS.

Figures (5)

• Figure 1: Label noise of both event and occurrence time points. Different shapes represent different types of events, and the noise affects the type of event and the occurrence time. The dashed lines represent the influence of noise.
• Figure 2: (a) illustrates the training process of RDHP. The re-weighting net is frozen, outputting weights for re-weighting classification and regression losses. (b) depicts the re-weighting net. Before the training of the Hawkes model, the re-weighting net is trained using the losses from a clean dataset. Then, these parameters are frozen for RDHP training.
• Figure 3: Quantitative Impact of Different Noise Types on Deep Hawkes Process Intensity on MIMIC-II dataset. The X-axis represents noise type, while the Y-axis represents the intensity difference between clean and noisy datasets under the same random seed with 30% uniform noise.
• Figure 4: Label Noise Types Applied in the Experiment ($K=4$, $p=0.3$).
• Figure 5: Weight variation under 30% uniform noise experiment on MIMIC-II dataset.