Query on Port
Closed this issue · 0 comments
siabdullah4 commented
I am trying to replicate the findings of the Port paper, I've encountered variations in the results for the asynchronous baseline while working with the EMNIST dataset, specifically regarding the minimum_clients_aggregated and avg_training_time parameters. I'm curious to know if these parameters should remain consistent across all asynchronous baselines for all datasets. Furthermore, the specified partition size for the EMNIST dataset is 1128. What was the testset_size? Could you please provide insights or clarification on this matter?