https://github.com/jopasserat/federated-learning-tutorial

Hands-on part of the Federated Learning and Privacy-Preserving ML tutorial given at VISUM 2022
https://github.com/jopasserat/federated-learning-tutorial

differential-privacy dp-sgd federated-learning flower medical-imaging opacus ppml privacy-enhancing-technologies privacy-preserving-machine-learning

Last synced: 29 days ago
JSON representation

Hands-on part of the Federated Learning and Privacy-Preserving ML tutorial given at VISUM 2022

• Host: GitHub
• URL: https://github.com/jopasserat/federated-learning-tutorial
• Owner: jopasserat
• License: apache-2.0
• Created: 2022-07-08T23:59:00.000Z (over 2 years ago)
• Default Branch: master
• Last Pushed: 2022-07-22T11:18:32.000Z (over 2 years ago)
• Last Synced: 2024-12-25T18:30:29.254Z (about 1 month ago)
• Topics: differential-privacy, dp-sgd, federated-learning, flower, medical-imaging, opacus, ppml, privacy-enhancing-technologies, privacy-preserving-machine-learning
• Language: Python
• Homepage:
• Size: 116 KB
• Stars: 5
• Watchers: 3
• Forks: 1
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# FL simulation with medical imaging classification task

This code splits the Pathology MedMNIST dataset into `pool_size` partitions (user defined) and does a few rounds of training.

## Requirements

*    Flower 0.19.0

*    A recent version of PyTorch. This example has been tested with Pytorch 1.11.0

*    A recent version of Ray. This example has been tested with Ray 1.11.1.

### Install

Create a new Conda environment with Python 3.9, the following commands will isntall all the dependencies needed:

```

conda create --name my_project_env --file conda-linux-64.lock

poetry install

```

### Updating the environment

```

# Re-generate Conda lock file(s) based on environment.yml

conda-lock -k explicit --conda mamba

# Update Conda packages based on re-generated lock file

mamba update --file conda-linux-64.lock

# Update Poetry packages and re-generate poetry.lock

poetry update

```

## How to run

This example:

1. Downloads Pathology MedMNIST

2. Partitions the dataset into N splits, where N is the total number of

   clients. We refere to this as `pool_size`. The partition can be IID or non-IID

4. Starts a Ray-based simulation where a % of clients are sample each round.

   This example uses N=3, so 3 clients will be sampled each round.

5. After the M rounds end, the global model is evaluated on the entire testset.

   Also, the global model is evaluated on the valset partition residing in each

   client. This is useful to get a sense on how well the global model can generalise

   to each client's data.

The command below will assign each client 1 CPU threads. If your system does not have 1xN(=3) = 3 threads to run all 3 clients in parallel, they will be queued but eventually run. The server will wait until all N clients have completed their local training stage before aggregating the results. After that, a new round will begin.

```bash

$ python main.py --num_client_cpus 2 # note that `num_client_cpus` should be <= the number of threads in your system.

```

## References

- MedMNIST code adapted from this [Getting Started](https://github.com/MedMNIST/MedMNIST/blob/d8422ac64028488133fd21ff54372729e12bbaba/examples/getting_started.ipynb) example.

- Flower code adapted from this example: https://github.com/adap/flower/tree/2d45f12189984c2901d54e295f5c684b07039bd8/examples/simulation_pytorch