Deep Learning with Differential Privacy
CAIML #15 was an online meeting and took place on November 10. We had an excellent presentation by Kritika and a question & answer session, followed by an interactive networking session.
“A talk on how to train PyTorch models with differential privacy using the high-speed Opacus library. We discuss the basics of Differential Privacy, private Stochastic Gradient Descent and demonstrate the use of PyTorch’s Opacus library. We also compare it with Tensorflow’s privacy library.
Kritika leads the Differential Privacy Research Team at OpenMined. She is a Masters’ student at the Machine Learning Lab, IIIT Hyderabad. Her areas of interest are Deep Learning and Differential Privacy.”