Closed
Milestone
Jun 27, 2022–Sep 1, 2022
SP17 Item 05 -ensuring minimum number of samples before starting experimentation
Milestone ID: 2998
For some experiments, a requirement exists to only use datasets with at least a minumum number of samples. The goal is to prevent leakage of information by reconstructing the input data from the gradient updates.
The condition that must be respected:
- Each dataset used in an experiment by a participating site has a number of samples >= 20
We did not commit on the details about how we will enforce this.
Tasks:
-
[ ] bare-minimun: As a researcher, i want to check on the researcher side the number of samples in each dataset for my-> issue #316 (closed)Experiment
so that i can ensure they all have at leasts
samples-
minimal implementation: don't modifysee #316 (closed)Experiment()
, example code in notebook for checking this from experiment's datasets.
-
-
node sidesee #351
Validation of the task:
- Fed-BioMed team