Vector operations on vectors with different lenghts

Library Questions
1

Hello,

I have two vectors:

    vector<double> lat = {52.5212};
    vector<double> lats = {52.5115, 52.5122, 52.5125, 52.5121, 52.5195, 52.5132, 52.5105, 52.5047};

    auto subtraction = clientCC->EvalSub(ctLats, ctLat);

I want to subtract the value in lat from all values in lats homomorphically.

The results from the pseudocode above is:

(-0.0097, 52.5122, 52.5125, 52.5121, 52.5195, 52.5132, 52.5105, 52.5047, ... ); Estimated precision: 37 bits

The first element is obviously correct. My first thought was to create vector of same size like lats like:
vector<double> latConstants = {52.5212, 52.5212, 52.5212, 52.5212, 52.5212, 52.5212, 52.5212, 52.5212};

and then do the same EvalSub operation again. The problem is that lat is not existing in plaintext in the process. I only have Enc(lat) and i dont know how to get from Enc(lat) to Enc(latConstants)
What are your thoughts?

2

Please read about SIMD capabilities and homomorphic rotations, for instance, from the webinars or by searching in the forum.

If you have your ciphertext encrypting ct = [lat, 0, 0, … ], you can obtain an encryption of ct_rep = [lat, lat, lat, …], by repeatedly rotating and summing: ct_rep = ct + EvalAtIndex(ct, -1); ct_rep = ct_rep + EvalAtIndex(ct_rep, -2); etc (it is you can also do it with a linear number of rotations but it is more efficient to do it with a logarithmic number of rotations). This is done internally by the EvalSum() function.

3

Thank you very much! Your information was very useful. Espacially the way how the EvalSum() function works =)