[BUG] RECI implementation flaw?

[baosws]

I notice that in the current implementation (cdt 0.5.23) of the RECI algorithm there are these two lines (b_fit_score function):

poly_x[:,1] = 0
poly_x[:,2] = 0

which essentially supress the coefficients of x and x^2 in the polynomial. For the default degree of 3, this is basically fitting the function f(x) = a * x^3 + b in both causal directions.

I am pretty sure this is not mentioned in the original paper.

Is it intended for some reason? Please clarify.

Thank you!

[diviyank]

Hello,

I will look into it, it seems strange indeed