< 0.01, and 75th-percentiles of the total number of gradient descent steps used (across all networks G

Open in new window