Relating Graph Neural Networks to Structural Causal Models

Zečević, Matej; Dhami, Devendra Singh; Veličković, Petar; Kersting, Kristian

doi:10.48550/arxiv.2109.04173

Cited by 8 publications

(12 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…in the political or medical domain. We thus encourage anyone who uses VACA (or any other ML method for causal inference) to i) fully understand the model assumptions and to verify (up to the possible extend) that they are fulfilled; as well as ii) to be aware of the identifiability problem in counterfactual queries [49,45].…”

Section: Conclusion Limitations and Impactmentioning

confidence: 99%

See 1 more Smart Citation

VACA: Design of Variational Graph Autoencoders for Interventional and Counterfactual Queries

Sánchez‐Martín¹,

Rateike²,

Valera³

2021

Preprint

View full text Add to dashboard Cite

In this paper, we introduce VACA, a novel class of variational graph autoencoders for causal inference in the absence of hidden confounders, when only observational data and the causal graph are available. Without making any parametric assumptions, VACA mimics the necessary properties of a Structural Causal Model (SCM) to provide a flexible and practical framework for approximating interventions (do-operator) and abduction-action-prediction steps. As a result, and as shown by our empirical results, VACA accurately approximates the interventional and counterfactual distributions on diverse SCMs. Finally, we apply VACA to evaluate counterfactual fairness in fair classification problems, as well as to learn fair classifiers without compromising performance.

show abstract

Section: Conclusion Limitations and Impactmentioning

confidence: 99%

“…We stress that the causal graph can often be inferred from expert knowledge [52] or via one of the approaches for causal discovery [12,42]. With this analysis we aim to complement the concurrent line of research that theoretically studies the use of Neural Networks (NN) [45], and more recently GNNs [49], for causal inference.…”

Section: Introductionmentioning

confidence: 99%

VACA: Design of Variational Graph Autoencoders for Interventional and Counterfactual Queries

Sánchez‐Martín¹,

Rateike²,

Valera³

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Recent work [6] explores how to select trustworthy neighbors for GNN in the inference stage, and demonstrates its effectiveness in node classification tasks. [37] studies the connection between GNNs and SCM from a theoretical perspective. Different from them, we introduce a causal intervention strategy to mitigate the confounding effect for GNNs in the training stage.…”

Section: Causal Inferencementioning

confidence: 99%

Causal Attention for Interpretable and Generalizable Graph Classification

Sui,

Wang,

et al. 2021

Preprint

View full text Add to dashboard Cite

Learning powerful representations is one central theme of graph neural networks (GNNs). It requires refining the critical information from the input graph, instead of the trivial patterns, to enrich the representations. Towards this end, graph attention and pooling methods prevail. They mostly follow the paradigm of "learning to attend". It maximizes the mutual information between the attended subgraph and the ground-truth label. However, this training paradigm is prone to capture the spurious correlations between the trivial subgraph and the label. Such spurious correlations are beneficial to in-distribution (ID) test evaluations, but cause poor generalization in the out-of-distribution (OOD) test data. In this work, we revisit the GNN modeling from the causal perspective. On the top of our causal assumption, the trivial information serves as a confounder between the critical information and the label, which opens a backdoor path between them and makes them spuriously correlated. Hence, we present a new paradigm of deconfounded training (DTP) that better mitigates the confounding effect and latches on the critical information, to enhance the representation and generalization ability. Specifically, we adopt the attention modules to disentangle the critical subgraph and trivial subgraph. Then we make each critical subgraph fairly interact with diverse trivial subgraphs to achieve a stable prediction. It allows GNNs to capture a more reliable subgraph whose relation with the label is robust across different distributions. We conduct extensive experiments on synthetic and real-world datasets to demonstrate the effectiveness. CCS CONCEPTS• Mathematics of computing → Graph algorithms; • Computing methodologies → Neural networks.

show abstract

“…The regularization term introduced in Conj.1 thereby penalizes CEM that would not be able to account for certain explanations to the posed single-why questions. Furthermore, Conj.1 poses an interesting direction of future research concerning more general classes of causal models that could enable a tighter integration between current practices in deep learning and causality, as in (Xia et al, 2021) for general Neural Causal Models (NCM) and (Zečević et al, 2021b) for Graph Neural Networks based NCM.…”

Section: Mathematical Foundations Of Structural Causal Interpretationsmentioning

confidence: 99%

“…Since the CEM is a sub-structure of the SCM as it "summarizes" the given structural equations to their causal effects for each of the edges within the SCM's induced graph, we can make the important observation that methods that in fact model the actual SCM (e.g. NCMs as in (Xia et al, 2021) or (Zečević et al, 2021b)) are more powerful than methods that model the sub-structure (CEM).…”

Section: A3 Proofs For Theorem 2 and Propositionmentioning

confidence: 99%

XAI Establishes a Common Ground Between Machine Learning and Causality

Zečević¹,

Dhami²,

Rothkopf³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Human mental processes allow for qualitative reasoning about causality in terms of mechanistic relations of the variables of interest, which we argue are naturally described by structural causal model (SCM). Since interpretations are being derived from mental models, the same applies for SCM. By defining a metric space on SCM, we provide a theoretical perspective on the comparison of mental models and thereby conclude that interpretations can be used for guiding a learning system towards true causality. To this effect, we present a theoretical analysis from first principles that results in a human-readable interpretation scheme consistent with the provided causality that we name structural causal interpretations (SCI). Going further, we prove that any existing neural induction method (NIM) is in fact interpretable. Our first experiment (E1) assesses the quality of such NIM-based SCI. In (E2) we observe evidence for our conjecture on improved sample-efficiency for SCI-based learning. After conducting a small user study, in (E3) we observe superiority in human-based over NIM-based SCI in support of our initial hypothesis.

show abstract

Relating Graph Neural Networks to Structural Causal Models

Cited by 8 publications

References 0 publications

VACA: Design of Variational Graph Autoencoders for Interventional and Counterfactual Queries

VACA: Design of Variational Graph Autoencoders for Interventional and Counterfactual Queries

Causal Attention for Interpretable and Generalizable Graph Classification

XAI Establishes a Common Ground Between Machine Learning and Causality

Contact Info

Product

Resources

About