Causality Based Learning
Causality-based learning aims to move beyond simple correlation, focusing on identifying and leveraging causal relationships within data to improve model performance and generalizability. Current research emphasizes integrating causal discovery methods, often using directed acyclic graphs (DAGs), with various machine learning architectures, including neural networks and generative adversarial networks (GANs), to build models that explicitly incorporate causal structure. This approach shows promise in enhancing prediction accuracy, particularly in handling out-of-distribution data and improving robustness across diverse applications, such as autonomous driving and conversational AI.
9papers
Papers
November 25, 2024
February 17, 2024
October 13, 2022
June 3, 2022
February 17, 2022