Unveiling the Sparsity Principle in Partially Observable Causal Representation Learning
The author establishes identifiability results for linear and piecewise linear mixing functions in a partially observed setting, emphasizing the importance of enforcing sparsity in representation learning.
Few-Shot Causal Representation Learning for Out-of-Distribution Generalization on Heterogeneous Graphs
Addressing out-of-distribution generalization challenges in heterogeneous graph few-shot learning through a causal model.
Heterogeneous Graph Few-Shot Learning for Out-of-Distribution Generalization
提案されたCOHFモデルは、構造的因果関係モデルを活用し、異なる分布間の分布シフトに対処するために無効な変数E2とZ2に焦点を当てています。
Identifiable Latent Neural Causal Models: Understanding Causal Representations Through Distribution Shifts
Understanding the identifiability of latent causal models through distribution shifts is crucial for predicting under unseen distributions.
Identifying Linearly-Mixed Causal Representations from Multi-Node Interventions in Causal Representation Learning
Learning causal representations from multi-node interventions using linear mixing can be achieved by exploiting sparsity in the variance of latent variables.
Causal Discovery with Single-Parent Decoding for Temporal Data: An Identifiable Causal Representation Learning Approach
Causal Discovery with Single-parent Decoding (CDSD) is a novel method for learning causal representations from temporal data by leveraging a sparsity assumption (single-parent decoding) to achieve identifiability of both the latent representation and the causal graph over these latents.
Identifiability of Latent Polynomial Causal Models with Changes in Causal Influences
This research paper presents a novel theoretical framework for identifying latent causal representations in polynomial causal models by leveraging changes in causal influences across multiple environments, generalizing previous work limited to linear Gaussian models.
Leveraging Pre-trained Language Models for Robust Causal Representation Learning in Single-Domain Scenarios
By leveraging the inherent domain shift between pre-trained and fine-tuned language models, we can construct robust causal representations that improve out-of-domain generalization in natural language understanding tasks, even with single-domain data.