PI-FL-Personalized-and-Incentivized-Federated-Learning

This is the code repository for the PI-FL: Personalized and Incentivized Federated Learning paper. Currently under submission at AAAI'23. Use of this code is allowed only for academic purposes.

How to set-up

Conda (recommended)

If you don't have Conda, you can install it here.

Once installed, create a new Conda environment. We recommend using Python 3.7, but newer versions may also work. For Mac M1/M2 systems, you must use Python 3.8 or above.

conda create -n <env_name> python=3.7

Activate the newly created Conda environment.

conda activate <env_name>

Setup the conda environment and IBMFL using the setup guide in

# Then replace the code in the conda environment /anaconda3/envs/PIFL/lib/python3.7/site-packages/ibmfl with PI-FL's code in this repository.

**Notes:**

* The quotes are required if using the Zsh shell (this is the default shell for Mac).
* There should be no spaces before or after each comma.
* The Keras backend will only work for Python 3.7; therefore, it will not work for Mac M1/M2 systems.

## Split Sample Data

You can use `generate_data.py` to generate sample data on any of the integrated datasets. This script requires the following flags:

| Flag | Description | Type |
| - | - | - |
| `-n <num_parties>` | the number of parties to split the data into | integer |
| `-d <dataset>` | which data set to use | string |
| `-pp <points_per_party>` | the number of data points per party | integer |

For example to generate data for **2 parties** with **200 data points** each from the **MNIST dataset**, you could run:

```sh
python examples/generate_data.py -n 2 -d mnist -pp 200

Run python examples/generate_data.py -h for full descriptions of the different options.

By default the data is scaled down to range between 0 and 1 and reshaped such that each image is (28, 28). For more information on what preprocessing was performed, check the Keras classifier example.

Create Configuration Files

To run IBM federated learning, you must have configuration files for the aggregator and for each party.

Before generating the configs edit the aggregator IP in line 80 in generate_configs.py

#On line 80, replace '192.168.0.231' with aggregator machine's IP.

You can generate these config files using the generate_configs.py script. This script requires the following flags:

Flag	Description	Type
-f <fusion>	which fusion algorithm to run	string
-m <model>	which framework model to use (sklearn, pytorch, keras, tf)	string
-n <num_parties>	the number of parties to split the data into	integer
-d <dataset>	which data set to use	string
-p <path>	path to load saved config data	string
-tn <int>	Number of tiers	int
--party-ip <IP address>	IP address of party's machine	IP address

The -n <num_parties> and -d <dataset> flags should be the same same as when generating the sample data. The -p <path> flag will depend on the generated data from the previous step, but will typically be -p examples/data/<dataset>/random. The backend framework for model from the -m <model> flag must be installed.

This script will generate config files as follows:

# aggregator config
examples/configs/<fusion>/<model>/config_agg.yml
# party configs
examples/configs/<fusion>/<model>/config_party0.yml
examples/configs/<fusion>/<model>/config_party1.yml
...
examples/configs/<fusion>/<model>/config_party<n-1>.yml

For example to generate the configs for a PyTorch model for 2 parties using the tiered federated average fusion algorithm from the CIFAR10 dataset (generated from before), you could run:

python examples/generate_configs.py -n 100 -tn 2 -d cifar -p examples/data/cifar/balanced --fusion fedavg --model pytorch --party_ip 192.168.0.232

This command will generate the following config files:

# aggregator config
examples/configs/iter_avg/pytorch/config_agg.yml
# party configs
examples/configs/iter_avg/pytorch/config_party0.yml
examples/configs/iter_avg/pytorch/config_party1.yml

Run python examples/generate_configs.py -h for full descriptions of the different options.

How does it work?

There is a docs folder with tutorials and API documentation to learn how to use and extend IBM federated learning. We also have a few video tutorials.

• Web Site
• Aggregator and party configuration tutorial
• API documentation

Citing PI-FL

If you use PI-FL, please cite the following reference paper:

@misc{khan2023pifl,
      title={PI-FL: Personalized and Incentivized Federated Learning}, 
      author={Ahmad Faraz Khan and Xinran Wang and Qi Le and Azal Ahmad Khan and Haider Ali and Jie Ding and Ali Butt and Ali Anwar},
      year={2023},
      eprint={2304.07514},
      archivePrefix={arXiv},
      primaryClass={cs.LG}
}

How to Run

First, run the aggregator using the following command:

"python -m ibmfl.aggregator.aggregator examples/configs///config_agg.yml"

Then you can use the partyManager to run the parties with multiprocessing, here is the command to run the first 50 parties:

"python partyManager/party_manager.py 50 > logs/log.txt"