epfml
evaluate libp2p as backend network stack
https://libp2p.io/
evaluate which bindings or deployments of libp2p would be most suitable for communicating gradients as in ML training, supporting mobile phone OSs, and for interfacing with e.g. PyTorch / jax or similar schemes
not an urgency now because we have the peer.js version, but we should come back to this issue sometime later, as libp2p offers many nice advantages such as elegant server-free peer discovery (using DHTs). we could use that later to discover peers who want to train the same task, similarly as the keyword based approach used here: https://github.com/dennis-tra/pcp
we should keep an eye on when the js version of libp2p would be ready to offer the (basic) functionality we need: https://github.com/libp2p/js-libp2p