Research map / Learning theory and algorithms / Prediction and decision losses
Learning from Aggregate Observations
Yivan Zhang, Nontawat Charoenphakdee, Zhenguo Wu, Masashi Sugiyama
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 |
|---|---|---|---|
| Use maximum-likelihood classification directly from triplet comparisons instead of triplet-loss representation learning plus clustering | Ours, triplet comparison · MNIST | accuracy: 22.77 -> 94.94 | Improved |
| Use maximum-likelihood classification directly from triplet comparisons instead of triplet-loss representation learning plus clustering | Ours, triplet comparison · FMNIST | accuracy: 27.27 -> 81.49 | Improved |
| Use maximum-likelihood classification directly from triplet comparisons instead of triplet-loss representation learning plus clustering | Ours, triplet comparison · KMNIST | accuracy: 20.39 -> 81.94 | Improved |
| Use maximum-likelihood classification directly from pairwise similarities instead of contrastive representation learning plus clustering | Ours, pairwise similarity · MNIST | accuracy: 98.45 -> 98.84 | Improved |
| Use maximum-likelihood classification directly from pairwise similarities instead of contrastive representation learning plus clustering | Ours, pairwise similarity · FMNIST | accuracy: 88.49 -> 90.59 | Improved |
| Use maximum-likelihood classification directly from pairwise similarities instead of contrastive representation learning plus clustering | Ours, pairwise similarity · KMNIST | accuracy: 89.65 -> 93.45 | Improved |
| Learn from group means using the aggregate likelihood instead of treating the mean as each instance's label | Ours, mean observation, GBM · abalone | MSE: 7.89 -> 4.80 | Improved |
| Learn from group means using the aggregate likelihood instead of treating the mean as each instance's label | Ours, mean observation, GBM · airfoil | MSE: 28.65 -> 4.63 | Improved |
| Learn from pairwise ranks with the Gaussian likelihood instead of RankNet's Gumbel model | Ours, rank observation, GBM · abalone | error variance: 10.66 -> 5.04 | Improved |
| Learn from pairwise ranks with the Gaussian likelihood instead of RankNet's Gumbel model | Ours, rank observation, GBM · airfoil | error variance: 47.46 -> 6.18 | Improved |
Closest papers by meaning
- A Universal Unbiased Method for Classification from Aggregate Observations (ICML 2023 · similarity 0.87)
- Model Agnostic Interpretability for Multiple Instance Learning (ICLR 2022 · similarity 0.86)
- PriorBoost: An Adaptive Algorithm for Learning from Aggregate Responses (ICML 2024 · similarity 0.85)
- Learning from Aggregate responses: Instance Level versus Bag Level Loss Functions (ICLR 2024 · similarity 0.85)
- Bipartite Ranking From Multiple Labels: On Loss Versus Label Aggregation (ICML 2025 · similarity 0.84)
- Probably Approximately Precision and Recall Learning (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.
