Research map / Optimization / Optimizers
Scalable and Effective Deep CCA via Soft Decorrelation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 exact decorrelation with mini-batch SDL | Deep CCA · MNIST, 50D correlation | correlation strength: 46.4 -> 45.5 | Hurt |
| Replace exact decorrelation with mini-batch SDL | Deep CCA · MNIST, 100D correlation | correlation strength: 89.5 -> 87.0 | Hurt |
| Replace exact decorrelation with mini-batch SDL | Deep CCA · MNIST, 500D correlation | correlation strength: 307.4 -> 356.8 | Helped |
| Replace exact decorrelation with mini-batch SDL | Deep CCA · Multi-PIE, correlation | correlation strength: 25.7 -> 29.2; 51.5 -> 60.5; 151.2 -> 163.2; 228.3 -> 257.7 | Helped |
| Add SDL decorrelation loss | FAE · MNIST | classification accuracy using z: 43.44 -> 11.35 | Helped |
| Add SDL decorrelation loss | FAE · MNIST | classification accuracy using y: 97.23 -> 97.33 | Helped |
| Add SDL decorrelation loss | 20-layer ResNet · CIFAR10 | accuracy (%): 91.12 -> 92.44 | Helped |
| Replace DeCov with SDL | 20-layer ResNet · CIFAR10 | accuracy (%): 91.62 -> 92.44 | Helped |
| Add SDL to each BN layer | DGDNet · Market-1501, single query | R1: 85.06 -> 86.75 | Helped |
| Add covariance-statistic accumulation to DeCov (ablation) | CNN classifier · CIFAR10 | accuracy (%): 91.62 -> 91.86 | Helped |
| Replace DeCov L2 formulation with L1 (ablation) | CNN classifier · CIFAR10 | accuracy (%): 91.62 -> 91.90 | Helped |
| Add covariance-statistic accumulation to DeCov (ablation) | DGDNet · Market-1501, single query | R1 accuracy (%): 85.86 -> 86.28 | Helped |
Closest papers by meaning
- Unconstrained Stochastic CCA: Unifying Multiview and Self-Supervised Learning (ICLR 2024 · similarity 0.86)
- Unpaired Canonical Correlation Analysis (NeurIPS 2026 · similarity 0.85)
- Factor Decorrelation Enhanced Data Removal from Deep Predictive Models (NeurIPS 2025 · similarity 0.84)
- Preventing Model Collapse in Deep Canonical Correlation Analysis by Noise Regularization (NeurIPS 2024 · similarity 0.84)
- On Sparse Canonical Correlation Analysis (NeurIPS 2024 · similarity 0.83)
- Feature learning from non-Gaussian inputs: the case of Independent Component Analysis in high dimensions (ICML 2025 · similarity 0.82)
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.