Newest papers

Causal Integrated Gradients: Scalable Feature Attribution Along Causal Paths

Joran Michiels, Anis Ismail, Maarten De Vos, Alejandro Sifrim · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.22126 · cited by 0 works (OpenAlex) · open-access PDF

Spanish

Estos métodos de atribución de caracteres basados en gradientes son los métodos de atribución de caracteres basados en gradientes existentes que representan dependencias de caracteres de datos que no escalan a altas dimensiones. Proponemos gradientes integrados causales (CIG) y gradientes esperados causales (CEG), que extenden gradientes integrados para incorporar la estructura causal anterior entre las características de entrada, mientras que mantienen su eficiencia computacional y, en condiciones en el gráfico causal, sus garantías axiomáticas. En muchos entornos, ya se sabe que un factor causal entre las características es ya conocido, por ejemplo, que un factor de transcripción regula causalmente la expresión de una estructura causal de un gen objetivo. Sin embargo, los métodos basados en gradientes existentes no pueden explotar esta estructura, sino

Machine translation (Meta M2M100); the English original follows.

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-$β$ 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.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Computer science, Causal model, Causal structure, Attribution, Scalability, Artificial intelligence

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange