Robust Fully Distributed Minibatch Gradient Descent with Privacy Preservation

Figure 4

Classification accuracy of the compressed gradient update on the data sets based on trace-based simulation. We vary key size (1024 or 2048) and maximum tree size (19 or 67).