Research map / Learning theory and algorithms / Prediction and decision losses
Bridging the Gap: Unifying the Training and Evaluation of Neural Network Binary Classifiers
Nathan Tsoi, Kate Candon, Deyuan Li, Yofti Milkessa, Marynel Vazquez
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 |
|---|---|---|---|
| train with soft-set F1 loss instead of BCE | neural binary classifier · Adult | F1-Score: 0.26 ± 0.06 -> 0.63 ± 0.02 | Improved |
| train with soft-set F1 loss instead of BCE | neural binary classifier · Mammography | F1-Score: 0.56 ± 0.11 -> 0.63 ± 0.04 | Improved |
| train with soft-set F1 loss instead of BCE | neural binary classifier · Kaggle | F1-Score: 0.50 ± 0.33 -> 0.83 ± 0.02 | Improved |
| train with soft-set F1 loss instead of BCE | neural binary classifier · CocktailParty | F1-Score: 0.70 ± 0.02 -> 0.75 ± 0.01 | Improved |
| soft-set F1 loss instead of adversarial F1 approach | neural binary classifier · CocktailParty | F1-Score: 0.30 ± 0.06 -> 0.75 ± 0.01 | Improved |
| optimize Accuracy with soft sets instead of BCE | neural binary classifier · CocktailParty | Accuracy: 0.85 ± 0.01 -> 0.85 ± 0.01 | No clear effect |
| optimize Accuracy with soft sets on imbalanced data | neural binary classifier · Mammography | F1-Score: 0.56 ± 0.11 -> 0.00 ± 0.00 | Hurt |
| optimize AUROC with soft sets instead of BCE | neural binary classifier · CocktailParty | AUROC: 0.78 ± 0.01 -> 0.57 ± 0.00 | Hurt |
| optimize AUROC via WMW statistic instead of BCE | neural binary classifier · Adult | AUROC: 0.58 ± 0.02 -> 0.50 ± 0.00 | Hurt |
| train with soft-set F1 loss instead of BCE | CNN · CIFAR-10-Frog | F1-Score: 0.59 ± 0.04 -> 0.73 ± 0.01 | Improved |
| train with soft-set Accuracy loss instead of BCE | CNN · CIFAR-10-Transportation | F1-Score: 0.88 ± 0.01 -> 0.92 ± 0.00 | Improved |
| increase beta in F-beta loss from 1 to 3 | neural binary classifier · Mammography | Recall: 0.55 ± 0.07 -> 0.81 ± 0.03 | Improved |
Closest papers by meaning
- Understanding the Loss Surface of Neural Networks for Binary Classification (ICML 2018 · similarity 0.87)
- Training Binary Neural Networks via Gaussian Variational Inference and Low-Rank Semidefinite Programming (NeurIPS 2024 · similarity 0.86)
- Error-Correcting Output Codes with Ensemble Diversity for Robust Learning in Neural Networks (AAAI 2021 · similarity 0.85)
- Classification with Deep Neural Networks and Logistic Loss (JMLR 2024 · similarity 0.85)
- The High-Dimensional Geometry of Binary Neural Networks (ICLR 2018 · similarity 0.85)
- On the Hardness of Training Deep Neural Networks Discretely (AAAI 2025 · similarity 0.83)
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.
