Public Funding for Corvus Research Limited

This project proposes an approach to solve a general class of supervised machine learning tasks via a differentially private version of Quantised Langevin Stochastic Dynamics (QLSD). This approach enables the trained function to satisfy a strong form of differential privacy, to cover a wide range of machine learning models, and to have an easily tunable parameter that provides an explicit trade-off between differential privacy, convergence performance and communication cost.