Research map / Label-efficient and robust learning / Recognition and long-tail
UnfoldML: Cost-Aware and Uncertainty-Based Dynamic 2D Prediction for Multi-Stage Classification
Yanbo Xu, Alind Khare, Glenn Matlin, Monish Ramadoss, Rishikesan Kamaleswaran, Chao Zhang, Alexey Tumanov
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use soft-gating instead of hard-gating (ablation) | UnfoldML · Sepsis-Septic Shock | approx. convex hull area: 17.58 -> 31.17 | Improved |
| use soft-gating instead of hard-gating (recommended config) (ablation) | UnfoldML · Sepsis-Septic Shock end-to-end | end-to-end AUC: 84.4 -> 88.8 | Improved |
| dynamic cascade instead of always using the most accurate model (UnfoldML-A) (ablation) | UnfoldML-S · Sepsis-Septic Shock end-to-end | end-to-end AUC: 87.2 -> 88.8 | Improved |
| dynamic cascade instead of always using the most accurate model (UnfoldML-A) (ablation) | UnfoldML-S · Sepsis-Septic Shock end-to-end | spatio-temporal cost ($): 311.8 -> 13.7 | Improved |
| use UnfoldML-S instead of cheapest-model-only pipeline (UnfoldML-C) (ablation) | UnfoldML-S · Sepsis-Septic Shock end-to-end | end-to-end AUC: 76.9 -> 88.8 | Improved |
| replace multi-class LSTM (Rec.) with UnfoldML-S | Multi-class LSTM · Sepsis-Septic Shock end-to-end | end-to-end AUC: 83.5 -> 88.8 | Improved |
| replace most accurate multi-class LSTM-A with UnfoldML-S | Multi-class LSTM-A · Sepsis-Septic Shock end-to-end | spatio-temporal cost ($): 269.0 -> 13.7 | Improved |
| replace most accurate multi-class LSTM-A with UnfoldML-S | Multi-class LSTM-A · Sepsis-Septic Shock end-to-end | end-to-end AUC: 88.9 -> 88.8 | Hurt |
| multi-stage cascade (ICK transition) for septic shock | UnfoldML-S · Septic shock early prediction | hours of early prediction: 24.0 -> 26.1 | Improved |
| replace LSTM-A with UnfoldML-S, single-stage sepsis | LSTM-A · Sepsis prediction | AUC: 92.8 -> 91.1 | Hurt |
| replace LSTM-A with UnfoldML-S, single-stage septic shock | LSTM-A · Septic Shock prediction | AUC: 96.9 -> 96.2 | Hurt |
| replace IDK-Cascade (1-D hard-gating) with UnfoldML-S soft-gating | IDK-Cascade · Septic Shock prediction | cost per call ($): 34.0 -> 10.0 | Improved |
Closest papers by meaning
- Systematic Comparison of Semi-supervised and Self-supervised Learning for Medical Image Classification (CVPR 2024 · similarity 0.83)
- OCCAM: Towards Cost-Efficient and Accuracy-Aware Classification Inference (ICLR 2025 · similarity 0.83)
- Risk Horizons: Structured Hypothesis Spaces for Longitudinal Clinical Prediction (NeurIPS 2026 · similarity 0.82)
- UniPro: Unified Multi-Mode Medical Image Segmentation from 2D Images to 3D Volumes via Propagation (NeurIPS 2026 · similarity 0.82)
- Novel Class Discovery in Chest X-rays via Paired Images and Text (AAAI 2024 · similarity 0.82)
- Dynamic Decision Learning: Test-Time Evolution for Abnormality Grounding in Rare Diseases (ICML 2026 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
