Research map / Probabilistic and scientific modeling / Causal discovery
CASTLE: Regularization via Auxiliary Causal Graph Discovery
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 |
|---|---|---|---|
| Use CASTLE instead of the strongest listed baseline regularizer | Supervised prediction model · Boston Housing | MSE: 0.131 ± 0.014 -> 0.123 ± 0.016 | Helped |
| Use CASTLE instead of MixUp | Supervised prediction model · Wine Quality | MSE: 0.712 ± 0.018 -> 0.708 ± 0.030 | Helped |
| Use CASTLE instead of SAE | Supervised prediction model · Facebook Metrics | MSE: 0.372 ± 0.168 -> 0.246 ± 0.153 | Helped |
| Use CASTLE instead of Input Noise | Supervised prediction model · Bioconcentration (regression) | MSE: 0.319 ± 0.022 -> 0.318 ± 0.036 | Helped |
| Use CASTLE instead of SAE | Supervised prediction model · Student Performance | MSE: 0.228 ± 0.022 -> 0.200 ± 0.020 | Helped |
| Use CASTLE instead of Dropout 0.2 | Supervised prediction model · Community | MSE: 0.327 ± 0.025 -> 0.326 ± 0.031 | Helped |
| Use CASTLE instead of MixUp | Supervised prediction model · Contraception Choice | AUROC: 0.778 ± 0.009 -> 0.787 ± 0.007 | Helped |
| Use CASTLE instead of MixUp | Supervised prediction model · Pima Diabetes | AUROC: 0.802 ± 0.016 -> 0.817 ± 0.004 | Helped |
| Use CASTLE instead of Input Noise | Supervised prediction model · Las Vegas Ratings | AUROC: 0.597 ± 0.041 -> 0.595 ± 0.032 | Hurt |
| Use CASTLE instead of Input Noise | Supervised prediction model · Statlog Heart | AUROC: 0.922 ± 0.005 -> 0.929 ± 0.007 | Helped |
| Use CASTLE instead of L1 | Supervised prediction model · Retinopathy | AUROC: 0.801 ± 0.013 -> 0.814 ± 0.014 | Helped |
| Select reconstruction targets using learned DAG neighbors rather than reconstructing all features | CASTLE versus SAE · Synthetic data with disconnected noise variables | average rank | Mixed |
Closest papers by meaning
- Quantifying Distributional Invariance in Causal Subgraph for IRM-Free Graph Generalization (NeurIPS 2025 · similarity 0.86)
- Effective Causal Discovery under Identifiable Heteroscedastic Noise Model (AAAI 2024 · similarity 0.86)
- Boosting Causal Structure Learning: An Asymmetric Exponential Modulation Gaussian-Based Adaptive Sample Reweighting Framework (AAAI 2025 · similarity 0.86)
- Sparse Causal Discovery with Generative Intervention for Unsupervised Graph Domain Adaptation (ICML 2025 · similarity 0.85)
- Differentiable Structure Learning with Ancestral Constraints (ICML 2025 · similarity 0.85)
- A Fixed-Point Approach for Causal Generative Modeling (ICML 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.