This repository aims to include trustworthy AI related projects from Huawei Noah's Ark Lab.
Current projects include:
-
Causal Structure Learning
-
Causal Disentangled Representation Learning
-
gCastle (or pyCastle, pCastle)
Welcome to participate in PCIC Causal Discovery Competition 2021 (June 8, 2021 ~ August 22, 2021). gCastle offers the competition module to be convenient for you to generate the submission file and evaluate your methods, and you can visit the submission section of the competition website to get the toy example.
- Causal_Discovery_RL: code, datasets, and training logs of the experimental results for the paper 'Causal discovery with reinforcement learning', ICLR, 2020. (oral)
- GAE_Causal_Structure_Learning: an implementation for 'A graph autoencoder approach to causal structure learning', NeurIPS Causal Machine Learning Workshop, 2019.
- Datasets:
- Synthetic datasets: codes for generating synthetic datasets used in the paper.
- Real datasets: a very challenging real dataset where the objective is to find causal structures based on time series data. The true graph is obtained from expert knowledge. We welcome everyone to try this dataset and report the result!
- We will also release the codes for other gradient-based causal structure learning methods.
- CausalVAE: code and datasets of the experimental results for the paper 'CausalVAE: Disentangled Representation Learning via Neural Structural Causal Models', CVPR, 2021. (accepted)
- Datasets: code for generating synthetic datasets used in the paper.
- This is a causal structure learning toolchain, which contains various functionality related to causal learning and evaluation.
- Most of causal discovery algorithms in gCastle are gradient-based, hence the name: gradient-based Causal Structure Learning pipeline.