kimandrik
commented
6 years ago Hello
Sorry for badly written code, I will try to explain the names:
cnum - is number of ciphertexts we need to pack the full dataset. wBits - is number of precision bits we have when encrypting the dataset. pBits - is number of precision bits we have for constants encoding (as constants encoding almost have no errors, we can make pBits smaller than wBits) lBits - is the number of bits above wBits for the final result (if result is bigger than 1, then we cannot fit it in wBits, we need more) aBits - is some parameter for sigmoid evaluation and normally is 2 or 3. kdeg - is degree of sigmoid approximation slots - number of slots in one ciphertext and is not bigger than N/2 sBits = log2(slots) from the code batch - how many features we pack in one ciphertext.
about encLGD, encMLGD, encNLGD: encNLGD works well. I think others work not so well, and maybe there are bugs in these methods, so just ignore them.
I'm learning your code, but I have some trouble. Please help me! In your code, I don't know the meaning of some parameters(such as, cnum, slots, sBits, bBits and so on). Where can I get the way to solve the problem? In your paper, I don't find the good way to know these parameters. some method(such as, encLGD, encMLGD, encNLGD), what is their different point?
Please help me, Thank you so much.