There is a slight discrepancy between the content of the paper and the code.

[githubdemima123456]

In Section 3.2, "Problem Setup and Notation," the variable Y is determined by the random variable Z1 and others, but in the corresponding code, the variable Y is determined by the random variable Z2 and others, and theoretically, it is also determined by Z2.

[kimmo1019]

Thanks for your feedback, you are right. (Z1, Z2) in the paper corresponds to (Z2, Z1) in the code. Will fix it in the next update

[kimmo1019]

For all users, we currently have no theoretical guidance for the choice of dim(Zi) so far without prior knowledge. In most of the experiments in our CausalEGM paper, Z1 Z2 have the same dimension. Our current new work is trying to mitigate this problem by providing more theoretical analysis.

[kimmo1019]

@githubdemima123456 Hi, please find the latest release v0.4.0, which solves the discrepancy between paper and code.