Federated Learning with Raspberry PI (PySyft)

We are a group of scholars in the study group PyTorch Robotics from the Secure and Private AI Scholarship Challenge by Facebook AI and Udacity working together to implement this tutorial by Daniele Gadler from OpenMined.

We will set up PySyft on two Raspberry Pis and learn how to train a Recurrent Neural Network on a Raspberry Pi via PySyft.

Purpose of the project

The purpose of using federated learning on a Raspberry Pi (RPI) is to build the model on the device so that data does not have to be moved to a centralized server. In addition to increased privacy, FL works well for Internet-of-Things applications because training can be done on the device instead of having to pass data between devices and a centralized server.

This project, which implements the OpenMined tutorial simulates the process using 2 RPIs to classify a person's surname with its most likely language of origin.

Read more about this here in the article written by Jess

Federated learning?

Would you like to know more about to federated learning? Look no further! Our team has prepared a few articles to get you up to speed:

Raspberry Pi?

Would you like to know what parts are needed and how to get started? Have a look at these articles written by the team:

Stuck on a problem?

Do not worry, we have it all covered for you. Head over to the troubleshooting section here

Implementations of the project

Have a look here to see the implementation of the project in different ways:

Process to run the project - Elena Kutanov
Process to run the project - Sergio Valderrama
PyTorch V1.0.0 Wheel file - Suparna S Nair
Implementation of the code and guide - Alejandro Ahumada
Implementation of the code using Nigerian surnames - Temitope Oladokun
Shivams journey - Shivam Raisharma
Implementing Federated learning on CIFAR10 Dataset - Ebinbin Ajagun
Classifying names with a character-level RNN - Nirupama Singh
Federated learning FashionMNIST - Sushil Ghimire
Federated learning with LSTM - Ivoline Ngong

Code
Find the code and dataset for the project here
Slides: Char RNN Names Classification - Check the slides here
What does it mean to train a Neural Network? Read article here

Shortcut to command line

Want a quick guide for the command line commands used? Have a look here

Topics
- PySyft
- Raspberry Pi

Project leaders

[x] Helena Barmer - @Helena Barmer
[x] Shashi Gharti - @Shashi Gharti

Contributors/Team

[x] Jess - @Jess
[x] Nirupama Singh - @Nirupama Singh
[x] Pooja Vinod - @Pooja Vinod
[x] Alex Ahumada - @Alex Ahumada
[x] Elena Kutanov - @Elena Kutanov
[x] Ayesha Manzur - @Ayesha Manzur
[x] Ivoline Ngong - @Ivy
[x] Joyce Chidiadi - @Joyce
[x] Temitope Oladokun - @Temitope Oladokun
[x] Shivam Raisharma - @Shivam Raisharma
[x] Sankalp Dayal - @Sankalp Dayal
[x] Sushil Ghimire - @Sushil Ghimire
[x] Juan Carlos Kuri Pinto - @Juan Carlos Kuri Pinto
[x] Ebinbin Ajagun - @Ebinbin Ajagun
[x] Suparna S Nair - @Suparna S Nair
[x] Sayed Maheen Basheer - @Sayed Maheen Basheer
[x] Sergio Valderrama - @Sergio Valderrama
[x] Stanislav Ladyzhenskiy - @Stanislav Ladyzhenskiy
[x] Mikaela Sanchez - @Mika
[x] Muhammad Naufil - @Muhammad Naufil
[x] cibaca - @cibaca

More information

For more articles, resources and links visit our wiki section here
Read this article about the shared experiences from three participants of the group: Virtual Team Experiences — Project Showcase Challenge
Future implementations of the project - Read our suggestions here
Jetson Nano: Setup for Federated Learning - Read more about this setup here
Would you like to learn more about Secure and Private AI? Join this free course from Udacity here
Join OpenMined Slack here
PyTorch Robotics logo designed by Alex Ahumada

shashigharti / federated-learning-on-raspberry-pi

readme