analiza aiqbench
This paper extends integrated gradients so that feature attributions follow a known causal graph, separating a feature's direct effect from effects routed through its descendants. It claims to keep the computational efficiency of gradient methods, which matters because existing dependency-aware attribution methods like causal Shapley values do not scale well. The evidence comes from a synthetic benchmark, simulated single-cell data, and one real dataset with only an approximate causal graph, so readers should weigh how much the method depends on the quality of the prior causal structure.
Causal Integrated Gradients: Scalable Feature Attribution Along Causal Paths
arXiv:2610.22126v1 Announce Type: cross Abstract: Existing feature attribution methods that account for feature dependencies do not scale to high dimensions. We propose causal integrated gradients (CIG) and causal expected gradients (CEG), which extend integrated gradients to incorporate prior causal structure among input features while retaining its computational efficiency and, under conditions on the causal graph, its axiomatic guarantees. In many settings, some causal structure between features is already known, for example, that a transcription factor causally regulates a target gene's expression. Yet existing gradient-based methods cannot exploit this structure, instead crediting a feature's downstream effects over its true upstream cause, such as ranking the target gene above the transcription factor that regulates it. Our key idea is to embed this causal structure by routing the integration path along the causal graph. The method thereby decomposes each feature's attribution into a direct effect and effects mediated through its causal descendants. We evaluate CIG in three settings of increasing realism. First, on a synthetic benchmark with a known causal model, CIG's attributions best match the ground-truth causal effect from Pearl's do-calculus, matching the accuracy of causal Shapley values at a fraction of the computational cost. Second, on simulated single-cell gene expression data with a known gene regulatory network (Dyngen), CIG most accurately recovers the transcription factors driving the simulation. Third, on real interferon-$\beta$ stimulation data, where only an approximate causal graph is available, CIG better prioritizes the transcription factors driving the measured response than DeepSHAP and integrated gradients. Together, these results show that incorporating causal structure into gradient-based attributions recovers upstream drivers that existing methods miss, without sacrificing scalability.
Abstract:Existing feature attribution methods that account for feature dependencies do not scale to high dimensions. We propose causal integrated gradients (CIG) and causal expected gradients (CEG), which extend integrated gradients to incorporate prior causal structure among input features while retaining its computational efficiency and, under conditions on the causal graph, its axiomatic guarantees. In many settings, some causal structure between features is already known, for example, that a transcription factor causally regulates a target gene's expression. Yet existing gradient-based methods cannot exploit this structure, instead crediting a feature's downstream effects over its true upstream cause, such as ranking the target gene above the transcription factor that regulates it. Our key idea is to embed this causal structure by routing the integration path along the causal graph. The method thereby decomposes each feature's attribution into a direct effect and effects mediated through its causal descendants. We evaluate CIG in three settings of increasing realism. First, on a synthetic benchmark with a known causal model, CIG's attributions best match the ground-truth causal effect from Pearl's do-calculus, matching the accuracy of causal Shapley values at a fraction of the computational cost. Second, on simulated single-cell gene expression data with a known gene regulatory network (Dyngen), CIG most accurately recovers the transcription factors driving the simulation. Third, on real interferon-$\beta$ stimulation data, where only an approximate causal graph is available, CIG better prioritizes the transcription factors driving the measured response than DeepSHAP and integrated gradients. Together, these results show that incorporating causal structure into gradient-based attributions recovers upstream drivers that existing methods miss, without sacrificing scalability.
41 pages (9 main text), 18 figures, 32 tables. Joran Michiels and Anis Ismail contributed equally. Code: this https URL
Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
arXiv:2610.22126 [cs.LG]
(or arXiv:2610.22126v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2610.22126 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 379 cuvinte
Subiecte conexe
PIRAMID: Progress and Plans
LessWrong · 27 iul. 2026
Evaluating the Causal Faithfulness of SHAP as a Candidate Attribution Method for SAE Features
arXiv · 15 oct. 2026
Why I Left Google DeepMind
LessWrong · 15 iul. 2026
Spectral Integrated Gradients for Coarse-to-Fine Feature Attribution
arXiv · 6 aug. 2026