Research map / Optimization / Optimizers
On Making Stochastic Classifiers Deterministic
Andrew Cotter, Maya Gupta, Harikrishna Narasimhan
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 |
|---|---|---|---|
| replace thresholding with hashing to derandomize the stochastic ensemble | stochastic classifier · Law School test, ROC matching | objective difference: 0.099 -> 0.001 | Improved |
| replace thresholding with hashing | stochastic classifier · COMPAS test, ROC matching | objective difference: 0.002 -> 0.005 | No clear effect |
| replace Average baseline with hashing for regression ensembles | stochastic regression ensemble · Law School test, histogram matching | squared-error difference: 0.262 -> 0.002 | Improved |
| replace hashing with variable binning (VarBin) | stochastic ensemble derandomization · ROC matching, histogram matching | objective difference | No clear effect |
| use VarBin instead of hashing at a small number of bins (ablation) | derandomized classifier · ROC matching (Figure 2) | objective difference vs bins | Improved |
| increase the number of bins in hashing (ablation) | hashing classifier · ROC matching (Figure 2) | accuracy on rate metrics vs orderliness | Mixed |
Closest papers by meaning
- Statistical Comparisons of Classifiers by Generalized Stochastic Dominance (JMLR 2023 · similarity 0.87)
- Optimal Binary Classification Beyond Accuracy (NeurIPS 2022 · similarity 0.86)
- On the Role of Randomization in Adversarially Robust Classification (NeurIPS 2023 · similarity 0.85)
- Individual Fairness In Strategic Classification (NeurIPS 2025 · similarity 0.84)
- Online Classification with Predictions (NeurIPS 2024 · similarity 0.83)
- Stochastic Gradients under Nuisances (NeurIPS 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.
