ptbfla

The package ptbfla comprises the following two FL frameworks:

1. Python TestBed for Federated Learning Algorithms (PTB-FLA) on a localhost.
2. MicroPython testbed for Federated Learning Algorithms (MPT-FLA) on a (W)LAN of PCs and IoTs like RPi Pico W boards.

Project team: Miroslav Popovic, Marko Popovic, Ivan Kastelan, Miodrag Djukic.

Project description:

1. For the introduction to PTB-FLA, including the examples 1, 2, and 3, see ZINC 2023 paper.
2. For the development paradigm, including the examples 4 and 5, see ECBS 2023 paper 1.
3. For the formal verification of PTB-FLA generic algorithms, see ECBS 2023 paper 2.
4. For the introduction to MPT-FLA, and the "mpasync" examples, see the paper and the video.

Directories (under the directory src):

• ptbfla_pkg: PTB-FLA (ptbfla.py, mpapi.py), MPT-FLA (mp_async_ptbfla.py, mp_async_mpapi.py), launcher.py (shared by both).
• examples: both PTB-FLA examples (without the prefix "mpasync") and MPT-FLA (with the prefix).
• examplesrp2_board: the modules config.py and main.py to install on RPi Pico W boards to run "mp_async" examples.

Notes on launching the examples:

• Remember to change the working directory to "src/examples".
• You must replace the IP addresses in the launch commands for the "mpasync" examples with your IP address!
• When running PTB-FLA apps on Windows with many instances e.g., ODTS for 170 satellites (launch example6_odts.py 170 id 1 25), Windows may start protecting itself by throwing the exception "WinError 10061" to some of the processes, this is as expected, PTB-FLA should be able to cope with this and make the recovery, you should not do anything but be patient and wait.

Warning: this project is work in progress.

Quick installation on Windows:

1. Clone the ptbfla project into the directory src
2. Create the venv: py -m venv venv_ptbfla
3. Activate the venv_ptbfla: venv_ptbfla\Scripts\activate
4. PIP-install the package ptbfla_pkg ("-e" means editable install): py -m pip install -e src
5. Run example 1: launch src\examples\example1_fedd_mean.py 3 id 0

Notes:

• If for some reason you want to uninstall the pacakage, type: pip uninstall venv_ptbfla
• To deactivate the venv_ptbfla on Windows type: venv_ptbfla\Scripts\deactivate.bat

Quick installation on Ubuntu:

1. Clone the ptbfla project into the directory src
2. Create the venv: python3 -m venv venv_ptbfla
3. Source the venv_ptbfla: source venv_ptbfla/bin/activate
4. PIP-install the package ptbfla_pkg ("-e" means editable install): python3 -m pip install -e src
5. Run example 1: launch src/examples/example1_fedd_mean.py 3 id 0

Note: To deactivate the venv_ptbfla on Ubuntu type: deactivate

Additional installation steps for Examples 4 and 5 on Windows:

1. If not already acive, activate the venv_ptbfla: venv_ptbfla\Scripts\activate
2. Upgrade PIP: py -m pip install --upgrade pip
3. PIP-install the package numpy: py -m pip install numpy
4. PIP-install the package pandas: py -m pip install pandas
5. PIP-install the package matplotlib: py -m pip install matplotlib
6. PIP-install the package scikit-learn: py -m pip install -U scikit-learn
7. Chage directory to examples: cd src\examples
8. Run example 4: launch example4_logistic_regression.py 3 id 2
9. Run example 5: launch example5_dec_log_regression.py 2 id

Additional installation steps for Examples 4 and 5 on Ubuntu:

1. If not already sourced, source the venv_ptbfla: source venv_ptbfla/bin/activate
2. Upgrade PIP: python3 -m pip install --upgrade pip
3. PIP-install the package numpy: python3 -m pip install numpy
4. PIP-install the package pandas: python3 -m pip install pandas
5. PIP-install the package matplotlib: python3 -m pip install matplotlib
6. PIP-install the package scikit-learn: python3 -m pip install -U scikit-learn
7. Chage directory to examples: cd src/examples
8. Run example 4: launch example4_logistic_regression.py 3 id 2
9. Run example 5: launch example5_dec_log_regression.py 2 id

Distributed Launcher for MPT-FLA

If you wish to make MPT-FLA application testing on multiple PCs easier, see the Distributed Launcher for MPT-FLA repository

PTB-FLA Babel adapter

If you wish to use PTB-FLA accross multiple devices, see the PTB-FLA plug in adapter for Babel repository