One shot GAN (OSGAN)

Context

This is the code for OSGAN paper(In preparation). OSGAN is an improved version of one shot federated learning.
Below are some one shot federated learning papers for your reference

1. Fusion Learning: A One Shot Federated Learning
2. Hybrid Fusion Learning: A Hierarchical Learning Model for Distributed Systems

🛠 Installation & Set Up

1. Prerequisites

   Python >=3.7
   Tensorflow-gpu = 2.2

2. Install using requirements.txt

   pip install -r requirements.txt

3. Create Conda Environment

   conda create --name <env> --file requirements.txt

4. Alternative easy option

   Create Azure datascience VM
   Goto predefined tensorflow environment using
   conda activate py37_tensorflow

🚀 Building and Running for results

1. Goto desired dataset folder

   cd dataset

2. For OSGAN IID results

   python3 osgan_mnist_iid.py

3. For OSGAN Non-IID results (applicable only for image datsets)

   python3 osgan_mnist_non_iid.py

4. For Federated IID and NonIID results (can edit in code file for IID or Non-IID)

   python3 federated_dataset.py

Folder structure

1. Plots

   Plots folder contains the generated plots for the paper 
   (Results are taken from corresponding folder)

2. Dataset

   Each dataset has a corresponding folder, where results are divided 
   based on clients, IID setup and the algorithm (OSGAN, Federated)

3. Fashion MNIST

   For Fashion MNIST dataset we have two folders, where one folder contains 
   the implementation of CGAN based OSGAN and other cantains GAN based OSGAN

4. Results

   Results folder in each setup wise results folder contains 
   information regarding testing accuracy and training accuracies