Research map / Probabilistic and scientific modeling / Variational and generative sampling
Learning Randomly Perturbed Structured Predictors for Direct Loss Minimization
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| learn the noise variance rather than fixing it to zero (ablation) | Direct Stochastic Learning · Bipartite matching, d=100 | Percentage of zero Prop. Any Wrong sequences: 74.9% -> 76.8% | Helped |
| learn the noise variance rather than fixing it to one (ablation) | Direct Stochastic Learning · Bipartite matching, d=100 | Percentage of zero Prop. Any Wrong sequences: 68.5% -> 76.8% | Helped |
| learn the noise variance rather than fixing it to zero (ablation) | Direct Stochastic Learning · CIFAR-10 test, k=3 | classification accuracy: 27% -> 90.7% | Helped |
| learn the noise variance rather than fixing it to one (ablation) | Direct Stochastic Learning · CIFAR-10 test, k=3 | classification accuracy: 89.95% -> 90.7% | Helped |
| learn the noise variance rather than fixing it to one (ablation) | Direct Stochastic Learning · MNIST test, k=3 | classification accuracy: 53.84% -> 99.4% | Helped |
| learn the noise variance rather than fixing it to zero (ablation) | Direct Stochastic Learning · MNIST test, k=3 | classification accuracy: 99.2% -> 99.4% | No clear effect |
| learn the noise variance rather than fixing it to zero (ablation) | Direct Stochastic Learning · Fashion-MNIST test, k=5 | classification accuracy: 93.5% -> 94% | Helped |
| learn the noise variance rather than fixing it to one (ablation) | Direct Stochastic Learning · Fashion-MNIST test, k=5 | classification accuracy: 93.3% -> 94% | Helped |
| use Direct Stochastic Learning instead of NeuralSort deterministic | k-nearest-neighbor image classifier · CIFAR-10 test, k=1 | classification accuracy: 88.7% -> 29.6% | Hurt |
| use Direct Stochastic Learning instead of NeuralSort deterministic | k-nearest-neighbor image classifier · CIFAR-10 test, k=3 | classification accuracy: 90.0% -> 90.7% | Helped |
Closest papers by meaning
- Learning latent variable structured prediction models with Gaussian perturbations (NeurIPS 2018 · similarity 0.88)
- A Consistent Regularization Approach for Structured Prediction (NeurIPS 2016 · similarity 0.87)
- Learning Maximum-A-Posteriori Perturbation Models for Structured Prediction in Polynomial Time (ICML 2018 · similarity 0.86)
- Structured Prediction with Projection Oracles (NeurIPS 2019 · similarity 0.85)
- Efficient Gradient Computation for Structured Output Learning with Rational and Tropical Losses (NeurIPS 2018 · similarity 0.85)
- On Inference Stability for Diffusion Models (AAAI 2024 · 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.