Gradient Descent with Linearly Correlated Noise: Theory and Applications to Differential Privacy Anastasia Koloskova

Neural Information Processing Systems 

We study gradient descent under linearly correlated noise.