Hello,
I’m trying to understand the difference between these two errors: Unfeasible noise constraint and No parameters found
They have very similar descriptions in the documentation but I encounter them in different situations.
For example: Unfeasible noise constraint is thrown when I’m trying to add an array of large size and No parameters found is thrown when the algorithm is complex (involves lots of computations) and the numbers are large.
Is there any specific property that is associated with it? like a certain level of complexity or PBS count? What actually triggers these error? Also, how to resolve them?
I am aware of fhe.refresh() but I’m still wondering if there is any other solution.
Sorry for the late answer. It’s the same kind of problem, even if the error message is spelled differently.
Without seing the Python code, it’s a bit complicated to help. Maybe you would need to help a bit the compiler, by adding fhe.refresh(). Indeed, as far as I know, the compiler would not add PBS, even if it would be the only way to have something which works: eg, if you have a very large sum y = \sum w_i x_i (where w_i are constant integers and x_i are encrypted values), and then do z = T[y], it would not work without you splitting like:
y_0 = \partial_sum w_i x_i
y_1 = \partial_sum w_i x_i
…
y_k = \partial_sum w_i x_i
then y’_i = fhe.refresh(y_i), then y = \sum y’_i, then z = T[y]
