Label Enhanced Event Detection with Collective Knowledge and Heterogeneous Graph

Wu, Gongqing; Lu, Zhenya; Miao, Zhuochun; Zhuo, Xingrui; Zhang, Zan

doi:10.1109/ickg55886.2022.00046

Cited by 2 publications

(9 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Accuracy Boost Accuracy (a) The performance of DIR (Wu et al 2022), a GNN with causal enhancement method, and Empirical Risk Minimization (ERM), as conventional GNN, on datasets devoid of confounders. ERM and DIR employ identical backbone architectures with consistent network sizes.…”

Section: Degree Of Correlationmentioning

confidence: 99%

“…Since then, a multitude of GNN variants have emerged (Wang et al 2022a;Fu, Zhao, and Bian 2022;Zhang et al 2022), each addressing specific challenges. In addition, there's an increasing interest in enhancing GNNs' ability to model causal relationships (Wu et al 2022), as GNNs with causal enhancement aim to incorporate causal inference into graph learning, leading to more reliable predictions.…”

Section: Related Work Graph Neural Networkmentioning

confidence: 99%

“…Recently, causal learning methods have also been widely used in graph neural networks . Such methods (Wu et al 2022;Chen et al 2022;Gao et al 2023) discovered potential laws in graph representation learning by studying causality in graph learning, and improved the completion effect of corresponding downstream tasks. We aim to develop a sound analytical approach to thoroughly analyze these methods.…”

Section: Causal Learningmentioning

confidence: 99%

“…We adopt ERM and ASAP (Ranjan, Sanyal, and Talukdar 2020) as foundational benchmarks for conventional GNNs. Additionally, for GNNs with causal enhancement, we pick DIR (Wu et al 2022), CIGA (Chen et al 2022), DISC (Fan et al 2022b) and RCGRL (Gao et al 2023) as our baseline methods. These methods adopt the same GNN backbone as ERM.…”

Section: Evaluationsmentioning

confidence: 99%

“…These methods aim to eliminate the influence of confounder within graph data, as confounder can create a false association between the cause and effect due to its correlation with both (Pearl 2002). Specifically, some approaches (Wu et al 2022;Fan et al 2022a) involve partitioning the training data into causal components and confounders, followed by separate processing to enable the model to disregard the confounders. Alternatively, other methods (Chen et al 2022;Gao et al 2023) aim to directly identify causal data or eliminate confounders to achieve the modeling of causal relationships.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Rethinking Causal Relationships Learning in Graph Neural Networks

Gao,

Yao,

et al. 2024

AAAI

View full text Add to dashboard Cite

Graph Neural Networks (GNNs) demonstrate their significance by effectively modeling complex interrelationships within graph-structured data. To enhance the credibility and robustness of GNNs, it becomes exceptionally crucial to bolster their ability to capture causal relationships. However, despite recent advancements that have indeed strengthened GNNs from a causal learning perspective, conducting an in-depth analysis specifically targeting the causal modeling prowess of GNNs remains an unresolved issue. In order to comprehensively analyze various GNN models from a causal learning perspective, we constructed an artificially synthesized dataset with known and controllable causal relationships between data and labels. The rationality of the generated data is further ensured through theoretical foundations. Drawing insights from analyses conducted using our dataset, we introduce a lightweight and highly adaptable GNN module designed to strengthen GNNs' causal learning capabilities across a diverse range of tasks. Through a series of experiments conducted on both synthetic datasets and other real-world datasets, we empirically validate the effectiveness of the proposed module. The codes are available at https://github.com/yaoyao-yaoyao-cell/CRCG.

show abstract

Section: Degree Of Correlationmentioning

confidence: 99%

Section: Related Work Graph Neural Networkmentioning

confidence: 99%

Section: Causal Learningmentioning

confidence: 99%

Section: Evaluationsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Rethinking Causal Relationships Learning in Graph Neural Networks

Gao,

Yao,

et al. 2024

AAAI

View full text Add to dashboard Cite

show abstract

Rethinking Dimensional Rationale in Graph Contrastive Learning from Causal Perspective

Ji,

Li,

et al. 2024

AAAI

View full text Add to dashboard Cite

Graph contrastive learning is a general learning paradigm excelling at capturing invariant information from diverse perturbations in graphs. Recent works focus on exploring the structural rationale from graphs, thereby increasing the discriminability of the invariant information. However, such methods may incur in the mis-learning of graph models towards the interpretability of graphs, and thus the learned noisy and task-agnostic information interferes with the prediction of graphs. To this end, with the purpose of exploring the intrinsic rationale of graphs, we accordingly propose to capture the dimensional rationale from graphs, which has not received sufficient attention in the literature. The conducted exploratory experiments attest to the feasibility of the aforementioned roadmap. To elucidate the innate mechanism behind the performance improvement arising from the dimensional rationale, we rethink the dimensional rationale in graph contrastive learning from a causal perspective and further formalize the causality among the variables in the pre-training stage to build the corresponding structural causal model. On the basis of the understanding of the structural causal model, we propose the dimensional rationale-aware graph contrastive learning approach, which introduces a learnable dimensional rationale acquiring network and a redundancy reduction constraint. The learnable dimensional rationale acquiring network is updated by leveraging a bi-level meta-learning technique, and the redundancy reduction constraint disentangles the redundant features through a decorrelation process during learning. Empirically, compared with state-of-the-art methods, our method can yield significant performance boosts on various benchmarks with respect to discriminability and transferability. The code implementation of our method is available at https://github.com/ByronJi/DRGCL.

show abstract

Label Enhanced Event Detection with Collective Knowledge and Heterogeneous Graph

Cited by 2 publications

References 27 publications

Rethinking Causal Relationships Learning in Graph Neural Networks

Rethinking Causal Relationships Learning in Graph Neural Networks

Rethinking Dimensional Rationale in Graph Contrastive Learning from Causal Perspective

Contact Info

Product

Resources

About