Research map / Learning theory and algorithms / Matrix and tensor methods
Learning Log-Determinant Divergences for Positive Definite Matrices
Anoop Cherian, Panagiotis Stanitsas, Jue Wang, Mehrtash Tafazzoli Harandi, Vassilios Morellas, Nikos Papanikolopoulos
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 dictionary jointly with alpha,beta instead of fixing alpha,beta (ablation) | IDDL · Virus (45 atoms) | accuracy: 80.33% -> 83.67% | Improved |
| learn alpha,beta in addition to the dictionary instead of fixing the dictionary to KMeans init (ablation) | IDDL · Virus (45 atoms) | accuracy: 70.00% -> 83.67% | Improved |
| learn alpha,beta jointly with dictionary instead of fixing alpha,beta (ablation) | IDDL · Virus (15 atoms) | accuracy: 78.33% -> 77.33% | Hurt |
| learn the dictionary instead of fixing it (ablation) | IDDL · Virus (15 atoms) | accuracy: 61.67% -> 77.33% | Improved |
| use structured-SVM objective instead of ridge regression | IDDL_N · Brodatz | accuracy: 77.1% -> 82.1% | Improved |
| use structured-SVM objective instead of ridge regression | IDDL_V · 3D Object | accuracy: 94.1% -> 98.3% | Improved |
| use structured-SVM objective instead of ridge regression | IDDL_V · Breast Cancer | accuracy: 90.0% -> 78.1% | Hurt |
| use structured-SVM objective instead of ridge regression | IDDL_V · KTH-TIPS2 | accuracy: 79.4% -> 71.3% | Hurt |
| learn alpha,beta per atom (IDDL-N) instead of a shared scalar pair (IDDL-S) (ablation) | IDDL · HMDB | accuracy: 52.30% -> 58.6% | Improved |
| use IDDL-Burg (alpha=beta=1) instead of learned alpha,beta per atom (ablation) | IDDL · 3D Object | accuracy: 92.57% -> 96.08% | Improved |
| use learned divergence IDDL instead of fixed AIRM divergence | IDDL · HMDB | accuracy: 43.20% -> 58.6% | Improved |
| use learned alpha-beta KMeans (alpha!=beta) instead of LE-KMeans | alpha-beta KMeans · Prostate Cancer clustering | F1-score: 0.578 -> 0.660 | Improved |
Closest papers by meaning
- Learning Discriminative ab-Divergences for Positive Definite Matrices (ICCV 2017 · similarity 0.95)
- Bayesian Nonparametric Clustering for Positive Definite Matrices (TPAMI 2016 · similarity 0.87)
- A Riemannian Network for SPD Matrix Learning (AAAI 2017 · similarity 0.87)
- Dimensionality Reduction on SPD Manifolds: The Emergence of Geometry-Aware Methods (TPAMI 2017 · similarity 0.87)
- Connecting Jensen–Shannon and Kullback–Leibler Divergences: A New Bound for Representation Learning (NeurIPS 2025 · similarity 0.85)
- Task-Induced Riemannian Metrics for Vision Transformer Feature Spaces (NeurIPS 2026 · 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.
