High Dimensional Distributed Gradient Descent with Arbitrary Number of Byzantine Attackers

Designing a new method for high-dimensional distributed learning under arbitrary Byzantine attackers, achieving minimax optimal statistical rates.