Research map / Learning theory and algorithms / Kernels and regression theory
Input Output Kernel Regression: Supervised and Semi-Supervised Structured Output Prediction with Operator-Valued Kernels
Céline Brouard, Marie Szafranski, Florence d'Alché-Buc
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 IOKR-ridge (least-squares loss) instead of IOKR-margin (hinge loss) | IOKR · synthetic networks (transductive) | AUC-PR | Improved |
| add the unlabeled-data smoothness constraint (semi-supervised, lambda2 > 0) | IOKR-margin and IOKR-ridge · NIPS co-authorship network, 2.5% labeled | AUC-ROC: up to 0.14 points for IOKR-margin and up to 0.11 points for IOKR-ridge | Improved |
| use IOKR-ridge instead of IOKR-margin | IOKR · NIPS co-authorship network | AUC-ROC and AUC-PR | Improved |
| use IOKR instead of EM at 5% labeled data | IOKR · NIPS co-authorship network (transductive) | AUC-ROC: 83.6 ± 5.9 vs 87.3 ± 2.4 | Hurt |
| use IOKR instead of PKMR at 20% labeled data | IOKR · NIPS co-authorship network (transductive) | AUC-PR: 43.7 ± 1.9 vs 38.8 ± 2.0 | Improved |
| use semi-supervised instead of transductive IOKR | IOKR · NIPS co-authorship network, 20% labeled | AUC-PR: 43.7 ± 1.9 -> 25.3 ± 3.0 | Hurt |
| use IOKR-ridge instead of OK3+ET | IOKR-ridge · Yeast PPI, integrated kernel | AUC-ROC: 86.9 ± 1.6 -> 91.2 ± 1.2 | Improved |
| use IOKR-ridge instead of OK3+ET | IOKR-ridge · Yeast PPI, integrated kernel | AUC-PR: 26.8 ± 2.4 vs 22.2 ± 1.6 | Hurt |
| use EM instead of IOKR at 5% labeled data | IOKR · Yeast PPI (transductive) | AUC-ROC: 80.6 ± 0.7 vs 82.2 ± 0.6 | Hurt |
| use semi-supervised instead of transductive IOKR | IOKR-ridge · Yeast PPI link prediction | AUC-ROC: 83.1 ± 0.5 vs 82.9 ± 1.2 | No clear effect |
| increase mu from 0 to 0.8 or 1 (use output relationships) | IOKR-ridge multi-task · NCI cancer cell line activity | MSE | Improved |
| add unlabeled data in multi-task IOKR | models 1 and 2 · NCI cancer cell line activity | MSE | Improved |
Closest papers by meaning
- Generalization Properties of hyper-RKHS and its Applications (JMLR 2021 · similarity 0.87)
- On the Target-kernel Alignment: a Unified Analysis with Kernel Complexity (NeurIPS 2024 · similarity 0.86)
- Generalized Tensor Decomposition for Understanding Multi-Output Regression under Combinatorial Shifts (NeurIPS 2024 · similarity 0.85)
- Generalization in Kernel Regression Under Realistic Assumptions (ICML 2024 · similarity 0.85)
- Spectrally Transformed Kernel Regression (ICLR 2024 · similarity 0.85)
- A Comprehensive Analysis on the Learning Curve in Kernel Ridge Regression (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.
