Research map / Probabilistic and scientific modeling / Causal discovery
Order-based structure learning for zero-inflated count data under data heterogeneity
Fariha Taskin, Junsouk Choi, Hyunwoong Chang
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes an order-based structure learning method for DAGs over zero-inflated multivariate count data, replacing local DAG-space search with an ordering-based scoring framework.
Problem
Causal structure learning for DAG models of zero-inflated multivariate count data; existing zero-inflated Poisson structure learners rely on local search over DAG space, which scales poorly and converges to suboptimal solutions.
Approach
Shifts the search from local DAG-space moves to an order-based scoring framework for zero-inflated Poisson models, exploiting identifiability of ZIP causal models and handling data heterogeneity.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| jointly estimate K=5 datasets instead of K=1 under a shared ordering (ablation) | ZIO · Simulation p=40 | Kendall tau: 0.311 -> 0.740 | Improved |
| increase K from 1 to 20 (ablation) | ZIO · Simulation p=40 | edge difference: 7.967 -> 4.062 | Improved |
| joint estimation with K=5 at 75% zeros, n=250 (ablation) | ZIO · Simulation p=20 | TPR: 0.597 -> 0.662 | Improved |
| use order-based search instead of DAG-space search | ZIO vs ZIPBN · Simulation p=75 | FDR: 0.281 -> 0.139 | Improved |
| order-based method vs ZICO at small p | ZIO vs ZICO · Simulation p=25 | TPR: 0.938 vs 0.917 | Hurt |
| increase zero proportion from 25% to 75% (ablation) | ZIO · Simulation p=20, n=250, K=1 | TPR: 0.903 -> 0.597 | Hurt |
| increase sample size from 250 to 2000 at 75% zeros (ablation) | ZIO · Simulation p=20, K=1 | TPR: 0.597 -> 0.937 | Improved |
| replace exact with forward-backward parent selection (ablation) | ZIO · Exhaustive p=7 | optimal recovery: 86.0 -> 84.0 (n=500); 94.0 / 94.0 (n=1000) | No clear effect |
| use five random starts instead of one (ablation) | R2R hill climbing · Exhaustive p=7, n=250 | finite-search success: 76.3 -> 95.5 | Improved |
| use adjacent transposition instead of R2R (ablation) | hill climbing · Simulation p=10, K=1 | penalized score convergence | Hurt |
| use random transposition instead of R2R (ablation) | hill climbing · Simulation p=10, K=1 | penalized score convergence | No clear effect |
| violate common ordering (U<=0.7) in joint estimation (ablation) | ZIO K=5 · Simulation p=10, n=1000 | Hamming distance | Hurt |
Experimental setup
Builds on: zero-inflated Poisson causal models
Closest papers by meaning
- Model-based Causal Discovery for Zero-Inflated Count Data (JMLR 2023 · similarity 0.90)
- Bayesian Causal Structural Learning with Zero-Inflated Poisson Bayesian Networks (NeurIPS 2020 · similarity 0.88)
- Boosting Causal Structure Learning: An Asymmetric Exponential Modulation Gaussian-Based Adaptive Sample Reweighting Framework (AAAI 2025 · similarity 0.88)
- Beware of the Simulated DAG! Causal Discovery Benchmarks May Be Easy to Game (NeurIPS 2021 · similarity 0.87)
- Learning Gaussian DAGs from Network Data (JMLR 2024 · similarity 0.87)
- Causal Discovery from Poisson Branching Structural Causal Model Using High-Order Cumulant with Path Analysis (AAAI 2024 · similarity 0.86)
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.
