Hello.
I have some questions about the encoding method in the CKKS scheme.
I feel that there are some gaps between the original encoding mechanism and the OpenFHE implementations.
Thus, I want to verify if my interpretation is correct and understand the reason for such implementations.
First, I am considering the following condition:
Then, my interpretation of the encoding implemented in the OpenFHE library (ckkspackedencoding.cpp) is as follows:
Is my interpretation roughly correct?
In the original paper,
I understood that the interpolation through inverse FFT uses not only the N/2 root of unity but also the N/2 conjugated root of unity.
Based on my understanding of the implementation, wasn’t the original paper intending to make P(X) a 2N-1 degree polynomial instead of an N/2-1 degree polynomial?
(It seems that the encoding implementations do not include the conjugated root of unity.)
My last question is about the reason for the implementation.
Why should the imaginary parts of IFFT coefficients b_l be used for generating P’(X)?
I have difficulty understanding the process of creating a polynomial with real coefficients (not complex number) through the inverse FFT.
( Or is it just for tracking the precision? Without the need for calculating precision, can I generate P’(X) as an N/2-1 degree polynomial with integer coefficients?)
Thank you for reading.