Research map / Learning theory and algorithms / Prediction and decision losses
Simultaneous Learning of Trees and Representations for Extreme Classification and Density Estimation
Yacine Jernite, Anna Choromanska, David Sontag
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 |
|---|---|---|---|
| Learn the tree structure jointly with representations instead of using a Huffman tree | FastText-style hierarchical classifier · YFCC100M, embedding dimension 50 | P@1: 28.3 -> 31.6; 29.9 -> 32.1 | Improved |
| Learn the tree structure jointly with representations instead of using a Huffman tree | FastText-style hierarchical classifier · YFCC100M, embedding dimension 200 | P@1: 35.8 -> 36.1; 36.4 -> 36.6 | Mixed |
| Increase tree arity from 2 to 5 | FastText-style hierarchical classifier · YFCC100M, embedding dimension 50 | P@1: 27.2 -> 28.3 | Improved |
| Increase tree arity from 5 to 20 | FastText-style hierarchical classifier · YFCC100M | P@1: 28.3 -> 29.9; 35.8 -> 36.4 | Improved |
| Use the learned tree rather than a random tree | Hierarchical language model · Gutenberg | perplexity: 160 -> 148 | Improved |
| Use the learned tree rather than a clustering-based tree | Hierarchical language model · Gutenberg | perplexity: 212 -> 148 | Improved |
| Use the learned hierarchical model rather than flat soft-max | Language model · Gutenberg | perplexity: 149 -> 148 | Improved |
| Use the learned hierarchical model rather than flat soft-max | Language model · Gutenberg | train ms/batch: 12.5 -> 4.5 | Improved |
| Use the learned hierarchical model rather than flat soft-max | Language model · Gutenberg | test ms/batch: 6.9 -> 0.9 | Improved |
Closest papers by meaning
- Learning Structured Representations by Embedding Class Hierarchy (ICLR 2023 · similarity 0.86)
- Self-Paced Unified Representation Learning for Hierarchical Multi-Label Classification (AAAI 2024 · similarity 0.86)
- GRANDE: Gradient-Based Decision Tree Ensembles for Tabular Data (ICLR 2024 · similarity 0.85)
- Joint Hierarchical Representation Learning of Samples and Features via Informed Tree-Wasserstein Distance (NeurIPS 2025 · similarity 0.85)
- Generative Forests (NeurIPS 2024 · similarity 0.85)
- Conditional Density Estimation with Histogram Trees (NeurIPS 2024 · similarity 0.85)
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.
