Federated learning using two augmentation techniques

Looking Federated learning experiment, the experiment contains 10 users, every user should have a four level fully connected CNN.

Three dataset should be used for the experiment CIFER 10, MINST and Kaggle MRI brain tumor.

The three dataset should be formatted as IID, non-IID and non-IID augmented with SMOTE and GAN techniques to balance the datasets, plot classes and number of samples for the 10 users for each different case.

Dataset plot output

Plot IID for 10 users and three datasets
Plot Non-IID for 10 users and three datasets
Plot non-IID-SMOTE for 10 users and three datasets
Plot non-IID-GAN for 10 users and three datasets


Apply federated learning using FedAVG with 100 iteration in each case and plot accuracy.



Plot summary of trails

Apply federated learning 100 iteration with IID (CIFER 10, MINST and Kaggle MRI)
Apply federated learning 100 iteration with non-IID (CIFER 10, MINST and Kaggle MRI)
Apply federated learning 100 iteration with non-IID-SMOTE (CIFER 10, MINST and Kaggle MRI)
Apply federated learning 100 iteration with non-IID-GAN (CIFER 10, MINST and Kaggle MRI)


FedAVG, Number of iteration (100) and local model should be fixed for all trails.

Non-IID distribution can be (5%,10%,20%,30%,40%,50%,60%,70%,80%,90%) ,percentage can be randomly assign for each class for number of samples