Wenqin Liu

Wenqin Liu

she/her

PhD Candidate, University of Melbourne. My research focuses on Causal Discovery and Machine Learning for robust and identifiable models.

Publications

Conference Papers

MissScore: High-Order Score Estimation in the Presence of Missing Data

Published in ICML, 2025

MissScore is a novel algorithm that enables causal discovery with incomplete datasets by leveraging high-order score function estimation, achieving state-of-the-art results in simulations.

Recommended citation: Wenqin Liu, Haonan Hou, Erdun Gao, Biwei Huang, Qiuhong Ke, Howard Bondell, Mingming Gong. "MissScore: High-Order Score Estimation in the Presence of Missing Data." ICML 2025.
Download Paper

A Skewness-Based Criterion for Addressing Heteroscedastic Noise in Causal Discovery

Published in ICLR, 2025

SkewScore introduces a skewness-based criterion to distinguish causal from anti-causal directions in models with heteroscedastic symmetric noise, and shows strong empirical and theoretical performance.

Recommended citation: Yingyu Lin*, Yuxing Huang*, Wenqin Liu*, Haoran Deng*, Ignavier Ng, Kun Zhang, Mingming Gong, Yi-An Ma, Biwei Huang. "A Skewness-Based Criterion for Addressing Heteroscedastic Noise in Causal Discovery." ICLR 2025.
Download Paper

Causal Discovery with Mixed Linear and Nonlinear Additive Noise Models: A Scalable Approach

Published in CLeaR, 2024

This paper proposes a causal discovery algorithm that identifies edge directions beyond Markov equivalence classes using the Jacobian of the score function, suitable for mixed linear and nonlinear mechanisms.

Recommended citation: Wenqin Liu, Biwei Huang, Erdun Gao, Qiuhong Ke, Howard Bondell, Mingming Gong. "Causal Discovery with Mixed Linear and Nonlinear Additive Noise Models: A Scalable Approach." CLeaR 2024.
Download Paper