Temporal Label Smoothing for Early Event Prediction

Hugo Yèche, Alizée Pace, Gunnar Ratsch, Rita Kuznetsova
Proceedings of the 40th International Conference on Machine Learning, PMLR 202:39913-39938, 2023.

Abstract

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.

Cite this Paper

BibTeX
@InProceedings{pmlr-v202-yeche23a, title = {Temporal Label Smoothing for Early Event Prediction}, author = {Y\`{e}che, Hugo and Pace, Aliz\'{e}e and Ratsch, Gunnar and Kuznetsova, Rita}, booktitle = {Proceedings of the 40th International Conference on Machine Learning}, pages = {39913--39938}, year = {2023}, editor = {Krause, Andreas and Brunskill, Emma and Cho, Kyunghyun and Engelhardt, Barbara and Sabato, Sivan and Scarlett, Jonathan}, volume = {202}, series = {Proceedings of Machine Learning Research}, month = {23--29 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v202/yeche23a/yeche23a.pdf}, url = {https://proceedings.mlr.press/v202/yeche23a.html}, abstract = {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.} }
Endnote
%0 Conference Paper %T Temporal Label Smoothing for Early Event Prediction %A Hugo Yèche %A Alizée Pace %A Gunnar Ratsch %A Rita Kuznetsova %B Proceedings of the 40th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2023 %E Andreas Krause %E Emma Brunskill %E Kyunghyun Cho %E Barbara Engelhardt %E Sivan Sabato %E Jonathan Scarlett %F pmlr-v202-yeche23a %I PMLR %P 39913--39938 %U https://proceedings.mlr.press/v202/yeche23a.html %V 202 %X 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.
APA
Yèche, H., Pace, A., Ratsch, G. & Kuznetsova, R.. (2023). Temporal Label Smoothing for Early Event Prediction. Proceedings of the 40th International Conference on Machine Learning, in Proceedings of Machine Learning Research 202:39913-39938 Available from https://proceedings.mlr.press/v202/yeche23a.html.

Related Material