Claudia Soares

Claudia Soares

I use optimization, physics, and probability theory to develop robust, interpretable, and trustworthy learning methods to be applied in the real world.

Federated and decentralized learning

My work in federated and decentralized learning asks how models can be trained when data, labels, features, or compute are split across institutions and devices. I am especially interested in methods that reduce communication without losing convergence guarantees, and in decentralized formulations where coordination is limited but the system still needs reliable learning or inference.

This line of work connects optimization theory with practical constraints: compressed communication, error feedback, semi-decentralized coordination, split neural networks, and distributed inference over large graphical models.

Publications in this topic