Research map / Learning theory and algorithms / Kernels and regression theory
Kernel Sufficient Dimension Reduction and Variable Selection for Compositional Data via Amalgamation
Junyoung Park, Jeongyoun Ahn, Cheolwoo Park
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 |
|---|---|---|---|
| Replace log-ratio/zero-replacement pipeline with kernel SDR over amalgamated compositions (proposed method) | compositional variable selection · Synthetic microbiome data, zero proportions 10-70% | avg true variables selected (of 10): Proposed 8.46-9.26 vs coda-lasso 2.32-6.3 and selbal 0.78-4.22 | Improved |
| Use amalgamation (aggregating unselected taxa into a dummy) instead of sub-composition for dimension reduction | compositional variable selection · Synthetic microbiome data (n=200, p=100) | mean true positives (of 10): 7.06 vs 4.96-5.00 for coda-lasso variants | Improved |
| Log-ratio methods with zero-replacement strategy varied (0.5xmin vs 1sum vs gbm) (ablation) | coda-lasso and selbal · Synthetic microbiome data (n=200, p=100) | mean true positives (of 10): coda-lasso 0.5xmin 4.96, 1sum 5.00, gbm 3.84 at m=10 | Mixed |
| Increase zero proportion from 10% to 70% in synthetic data | proposed method · Synthetic microbiome data (n=500, p=100) | avg true variables selected (of 10): 9.26 at 10% zeros down to 8.46 at 70% | Hurt |
| Proposed amalgamation SDR vs log-ratio methods on BMI regression | compositional variable selection · BMI microbiome regression, estimated MSE | estimated MSE (lower is better): 28.87-28.99 vs selbal 32.91-34.64; similar to coda-lasso 29.29 at m=3 | Improved |
| Varying the number of selected variables m (5 to 40) for proposed method (ablation) | proposed method · Synthetic microbiome data (n=200, p=100) | true signal variables selected: Increasing probability of selecting true variables as m grows; reaches bound in most cases | Improved |
| Increase sample size from 200 to 1000 for proposed method (ablation) | proposed method · Synthetic microbiome data (p=100, m=10) | true variables selected (of 10): power increases and converges to 10 (Table 3 proposed 7.06 at n=200 to 9.5 at n=800-1000) | Improved |
| Selbal forward selection, sample size increased (ablation) | selbal · Synthetic microbiome data | true positives: does not increase with n | No clear effect |
Closest papers by meaning
- The Maximum Separation Subspace in Sufficient Dimension Reduction with Categorical Response (JMLR 2020 · similarity 0.89)
- Enhancing Sufficient Dimension Reduction via Hellinger Correlation (ICML 2024 · similarity 0.88)
- Kernel Methods for Radial Transformed Compositional Data with Many Zeros (ICML 2022 · similarity 0.88)
- Heterogeneous Sufficient Dimension Reduction and Subspace Clustering (ICML 2025 · similarity 0.84)
- Neural Networks Perform Sufficient Dimension Reduction (AAAI 2025 · similarity 0.83)
- A Unified Framework for Variable Selection in Model-Based Clustering with Missing Not at Random (NeurIPS 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.
