3 papers
cs.LG2026
Let Confidence Change, Not the Prediction: Prediction-Preserving Repair for Post-hoc Calibration
Daehwan Kim, Haejun Chung, Ikbeom Jang
Post-hoc calibration corrects reported confidence, yet a multiclass calibrator can also change the associated top-1 prediction. Accuracy captures only the net effect of these chang…
cs.LG2026
When, Where, and How: Adaptive Binning for Tabular Self-Supervised Learning
Daehwan Kim, Haejun Chung, Ikbeom Jang
Medical tabular data are ubiquitous in clinical research, but deep learning for tables remains underexplored because reliable labels often require costly expert adjudication, even…
cs.LG2024
Ordinal-Aware Calibration for Ordinal Classification
Daehwan Kim, Haejun Chung, Ikbeom Jang
Deep neural networks frequently produce overconfident, miscalibrated predictions. In ordinal classification, predictions must also adhere to a unimodal and order-consistent structu…