Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces a temporally localized version of weighted score-oriented loss (wSOL) tailored for event prediction, addressing the mismatch between traditional pointwise accuracy metrics and the temporal utility of alarms in operational systems. By incorporating temporal weights that prioritize near-event accuracy and adjust penalties based on alarm timing, the authors demonstrate that wSOL can be effectively optimized using back-propagation. The evaluation across three benchmark datasets reveals that wSOL significantly enhances performance in scenarios where the relevance of predictions is time-sensitive and not captured by conventional labels.
Temporal localization in loss functions can drastically enhance event prediction accuracy, outperforming traditional methods by aligning model training with real-world alarm utility.
Operational event-detection systems are rarely assessed by pointwise accuracy alone. In anomaly detection, changepoint detection, and warning systems, the utility of an alarm depends on its temporal position relative to an event. This produces a score-loss mismatch. Neural networks are commonly trained with classical loss functions, such as cross-entropy, whereas deployment decisions are obtained by thresholding network predictions, merging alarms through post-processing rules, and evaluating them with event-based metrics defined by detection windows and false-alarm costs. This paper studies a temporally localized specialization of weighted score-oriented loss (wSOL) for event prediction. Starting from score-oriented losses based on expected confusion matrices and from the weighted SOL framework of Marchetti et al., we consider temporal weights that discount near-event false positives and reduce false-negative penalties when an event is preceded by an admissible alarm. The resulting objective is differentiable with respect to the network predictions, and therefore can be optimized by back-propagation. It can be instantiated with balanced accuracy, true skill statistic, F1, critical success index, and related confusion-matrix scores. We evaluate the proposed approach by comparing cross-entropy, unweighted score-oriented loss, and wSOL on three benchmark datasets for time-series event prediction and detection. The results show that wSOL can improve performance when the evaluation utility is localized in time and is not already encoded by the pointwise labels.