A timestamp-attributed event graph with bidirectional edges predicts ICU mortality at AUROC 0.8449, but the paper's ablations show heterogeneous edge typing did not help and its flagship 'bidirectional necessity' claim conflicts with its own edge definitions.
Continuous Graph Neural Networks,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
CT-HEG: A Bidirectional, Timestamp-Attributed Event Graph for ICU In-Hospital Mortality Prediction - An Architectural Ablation Study
A timestamp-attributed event graph with bidirectional edges predicts ICU mortality at AUROC 0.8449, but the paper's ablations show heterogeneous edge typing did not help and its flagship 'bidirectional necessity' claim conflicts with its own edge definitions.