Research map / Learning theory and algorithms / Kernels and regression theory
Sampled Softmax with Random Fourier Features
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 |
|---|---|---|---|
| Use RF-softmax instead of Quadratic | Sampled softmax · AmazonCat-13k | P@1: 0.84 -> 0.87 | Helped |
| Use RF-softmax instead of Quadratic | Sampled softmax · Delicious-200k | P@1: 0.40 -> 0.41 | Helped |
| Use RF-softmax instead of Quadratic | Sampled softmax · WikiLSHTC | P@1: 0.57 -> 0.56 | Hurt |
| Use RF-softmax instead of Uniform sampling | Sampled softmax · AmazonCat-13k | P@1: 0.83 -> 0.87 | Helped |
| Use RF-softmax instead of Uniform sampling | Sampled softmax · Delicious-200k | P@1: 0.36 -> 0.41 | Helped |
| Use RF-softmax instead of Uniform sampling | Sampled softmax · WikiLSHTC | P@1: 0.47 -> 0.56 | Helped |
| Use RF-softmax instead of exact softmax-distribution sampling | Sampled softmax · AmazonCat-13k | P@1: 0.87 -> 0.87 | No clear effect |
| Use RF-softmax instead of exact softmax-distribution sampling | Sampled softmax · Delicious-200k | P@1: 0.42 -> 0.41 | Hurt |
| Use RF-softmax instead of exact softmax-distribution sampling | Sampled softmax · WikiLSHTC | P@1: 0.58 -> 0.56 | Hurt |
| Normalize embeddings instead of leaving them unnormalized (ablation) | Full softmax · PennTreeBank | validation perplexity: 126 -> 120 | Helped |
| Normalize embeddings instead of leaving them unnormalized (ablation) | Full softmax · AmazonCat-13k | P@1: 87% -> 87% | No clear effect |
| Increase the RFF map dimension D (ablation) | RF-softmax · PennTreeBank | validation perplexity | Helped |
Closest papers by meaning
- Sampled Estimators For Softmax Must Be Biased (NeurIPS 2025 · similarity 0.90)
- Adaptive Sampled Softmax with Kernel Based Sampling (ICML 2018 · similarity 0.90)
- Adaptive Sampling for Efficient Softmax Approximation (NeurIPS 2024 · similarity 0.88)
- Density-Softmax: Efficient Test-time Model for Uncertainty Estimation and Robustness under Distribution Shifts (ICML 2024 · similarity 0.85)
- Sampling Binary Data by Denoising through Score Functions (ICML 2025 · similarity 0.84)
- Understanding the Training Speedup from Sampling with Approximate Losses (ICML 2024 · similarity 0.84)
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.