Some questions about algorithm implementation

[Boyle-Coffee]

Execute me. I wanna implement the algorithm introduced in the paper Diagnosis of Autism Spectrum Disorder by Causal Influence Strength Learned from Resting-State fMRI Data. The algorithm process is as follow:

#

And the is

The steps of causal discovery in the paper is similar to the algorithm implemented in this Git repository, so I want to implement the algorithm based on the code in this repository. Although there are some comments in the code, but there are questions. I hope you can take the time to answer my questions, Thank you.

• I first found this Git repository in the paper Estimating feedforward and feedback effective connections from fMRI time series: Assessments of statistical methods when I tried to implement the two-step algorithm. But I found there is no any formula in this paper, not even a likelihood formula. So I want to know is there other paper contain the formula about the code.
• In the two step, we need to Initialize the by ICA with adaptive lasso. But is the function natural_grad_Adasize_Mask an implementation of the ICA with adaptive lasso?
• What is the function scorecond for? Is the y_psi equal the in the formula(1)? Why is y_psi calculated every 12 iterations?
• Is the function adaptive_size an implementation of AdaGrad?