Learning Representations Using Causal Invariance

Learning Representations Using Causal Invariance

Explore the concept of causal invariance in representation learning through this NYU ECE seminar featuring Leon Bottou from Facebook AI Research. Delve into topics such as the proxy problem, correlation environments, mixture coefficients, and adversarial domain adaptation. Gain insights into practical applications like hiring problems and observer self-supervised learning. Discover how causal invariance can enhance prediction accuracy across different environments and improve the robustness of AI models.