IFoA-FL-WP

IFoA Use case [Privacy preserving ML collaboration on claims modelling]. We use Flower.dev framework to federate workflow.

1. pip install flwr
2. start server: python3 IFoAserver.py [please see the configuration of FL run in IFoAserver.py ; later on we will separate configuration from server code]
3. start FL training participants, --partition=i i in range(9) identifies participant's data chunk eg. to start client0 we call: python3 'IFoA client [ Multilayer ] [freMTPL2freq].py' --partition=0
4. results are stored in dedicated folders ag_0 .... ag_9

Execution:

1. Global model training ( no FL loop ): python3 'IFoA client [ Multilayer ] [freMTPL2freq].py' --partition=-1 --if_FL=0

2. Partial model training ( assuming 10 participants) . Model training for participant = i i in range(0,9):

3. FL training ( assuming 10 participants).

a) Start FL server (make sure the ip adress used in the code is correct):

b) Start participant == 0 with a call (make sure you use right ip adress and port when connecting to the server):

( you need to start 10 such processes representing participant==i i in range(10) )

TEST RUN 2 PARTICIPANTS:

1. Dylan: python3 'IFoA client [ Multilayer ] [freMTPL2freq].py' --agent_id=3
2. Malgorzata python3 'IFoA client [ Multilayer ] [freMTPL2freq].py' --agent_id=5