Estimated precision

FHE Questions
1

I get some questions. I am not sure about estimated precision.
In exampes i set: scaleModSize = 50;
it means that i multiply value with 2^50 and by encrypting divide /2^50. It develops 3types of errors.CKKS RNS implementation has an approximation error around 7 bits smaller(from your Paper). But how can calculate this estimated precision.
In your paper you multiply scaleModSize with value after embedding transformation, also with polynom of message?

Estimated precision means that difference to precision=50 is lost information?

2

Can you please formulate your question more clearly? Do these other threads answer your question? Meaning of the bit precision, Is there a way to estimate the precision for a calculation?

1 Like