Misc.
-
: A controllable text-to-image generation model grounded in a causal view of concept learning. By formulating concept learning as a causal latent variable identification problem and guiding it with causal principles, we establish theoretical guarantees for recovering multimodal concepts under mild assumptions. This formulation leads to improved controllability and stronger generation performance.
-
: Implementation of Rank-based Latent Causal Discovery (RLCD), a versatile causal discovery algorithm that allows causally related latent variables. RLCD can identify the causal structure over both latent and observed variables under mild conditions. The repository also includes code for identifying linear causal effects based on the discovered structure.
-
: Ever wonder what your personality is and how your personality is related to other things? As a group of researchers interested in AI for scientific discovery, we are currently conducting research involving personality measures, physical features (palm print, height, weight, etc.), and demographical information. We hope you can join us by completing the following survey. You will get your personality testing result at the end of the survey, and we will also update you on the research findings as soon as they are clear. We appreciate your time, your curiosity, and your contribution!
-
Lecture Notes:
- [2023/05, CBMS] Foundations of Causal Graphical Models and Structure Discovery (with 10 lectures) given by Kun Zhang and Peter Spirtes: and are available.