Beware of the Simulated DAG! Causal Discovery Benchmarks May Be Easy To Game
arXiv:2102.13647
Abstract
Simulated DAG models may exhibit properties that, perhaps inadvertently, render their structure identifiable and unexpectedly affect structure learning algorithms. Here, we show that marginal variance tends to increase along the causal order for generically sampled additive noise models. We introduce varsortability as a measure of the agreement between the order of increasing marginal variance and the causal order. For commonly sampled graphs and model parameters, we show that the remarkable performance of some continuous structure learning algorithms can be explained by high varsortability and matched by a simple baseline method. Yet, this performance may not transfer to real-world data where varsortability may be moderate or dependent on the choice of measurement scales. On standardized data, the same algorithms fail to identify the ground-truth DAG or its Markov equivalence class. While standardization removes the pattern in marginal variance, we show that data generating processes that incur high varsortability also leave a distinct covariance pattern that may be exploited even after standardization. Our findings challenge the significance of generic benchmarks with independently drawn parameters. The code is available at https://github.com/Scriddie/Varsortability.
accepted at the 35th Conference on Neural Information Processing Systems (NeurIPS), 2021; cf. https://openreview.net/forum?id=wEOlVzVhMW_
References in corpus (7)
- A Transformational Characterization of Equivalent Bayesian Network Structures
- On the Identifiability of the Post-Nonlinear Causal Model
- Strong Completeness and Faithfulness in Bayesian Networks
- DAG-GNN: DAG Structure Learning with Graph Neural Networks
- DAGs with No Fears: A Closer Look at Continuous Optimization for Learning Bayesian Networks
- Differentiable Causal Discovery Under Unmeasured Confounding
- Unsuitability of NOTEARS for Causal Graph Discovery