Acceleration of TenSEAL with parallel computing

[dimk1]

Question

Can TenSEAL be accelerated on a cluster of parallel computers?

Further Information

I think of building a small cluster of Raspberry Pis working as a small parallel computing structure and work on toy examples with TenSEAL to see if accelerations can be achieved. Are there ways for TenSEAL to be benefited from parallel structures and how can we decompose the operations in a parallel-supporting format?

[MartinNoc]

Hello @dimk1, this would be very interesting! As far as i know, the only way to parallel TenSEAL is to parallel the outer computation, so for example for a neural network forward pass to pipeline the network and to parallel the computation of nodes within the same layer as much as possible, but this seems tough as we cannot access or write single ckksvector (for example) elements.

Any comments on that? Best, Martin