Research map / Learning theory and algorithms / Prediction and decision losses
Rankmax: An Adaptive Projection Alternative to the Softmax Function
Weiwei Kong, Walid Krichene, Nicolas Mayoraz, Steffen Rendle, Li Zhang
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 Softmax cross-entropy with Rankmax adaptive projection loss | multi-label retrieval model · MovieLens 20M | AP@10: 0.182 -> 0.210 | Improved |
| replace Softmax with Rankmax | multi-label retrieval model · MovieLens 20M | Accuracy: 0.342 -> 0.397 | Improved |
| replace Sparsemax with Rankmax | multi-label retrieval model · MovieLens 20M | R@100: 0.464 -> 0.492 | Improved |
| replace Softmax with Rankmax | multi-label retrieval model · MovieLens 100K | AP@10 / Accuracy / R@100: AP@10 0.147 -> 0.154; R@100 0.401 -> 0.387 | Mixed |
| replace Softmax with Rankmax | multi-label retrieval model · MovieLens 1B | R@1000: 0.0117 -> 0.0116 | No clear effect |
| add exponential temperature annealing to Softmax | Softmax loss · MovieLens 20M | retrieval metrics | No clear effect |
Closest papers by meaning
- Adaptive Reduced Rank Regression (NeurIPS 2020 · similarity 0.85)
- Adaptive Sampling for Efficient Softmax Approximation (NeurIPS 2024 · similarity 0.85)
- Breaking the Softmax Bottleneck via Learnable Monotonic Pointwise Non-linearities (ICML 2019 · similarity 0.85)
- Computational and Statistical Tradeoffs in Learning to Rank (NeurIPS 2016 · similarity 0.84)
- Multivariate Soft Rank via Entropy-Regularized Optimal Transport: Sample Efficiency and Generative Modeling (JMLR 2023 · similarity 0.84)
- Revisiting the Nyström Method for Improved Large-scale Machine Learning (JMLR 2016 · 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.
