Temporal Label Smoothing for Early Event Prediction
This addresses the problem of low false-alarm rates in medical decision support systems, offering a novel but incremental improvement over existing methods.
The paper tackles early event prediction by proposing Temporal Label Smoothing (TLS), a method that exploits temporal dependencies to improve performance, reducing missed events by up to a factor of two over baselines on benchmark tasks.
Models that can predict the occurrence of events ahead of time with low false-alarm rates are critical to the acceptance of decision support systems in the medical community. This challenging task is typically treated as a simple binary classification, ignoring temporal dependencies between samples, whereas we propose to exploit this structure. We first introduce a common theoretical framework unifying dynamic survival analysis and early event prediction. Following an analysis of objectives from both fields, we propose Temporal Label Smoothing (TLS), a simpler, yet best-performing method that preserves prediction monotonicity over time. By focusing the objective on areas with a stronger predictive signal, TLS improves performance over all baselines on two large-scale benchmark tasks. Gains are particularly notable along clinically relevant measures, such as event recall at low false-alarm rates. TLS reduces the number of missed events by up to a factor of two over previously used approaches in early event prediction.