/v6-starter-crosstabulation-py

Crosstabulation algorithm, specific to STARTER project

Primary LanguagePythonMIT LicenseMIT


vantage6

A privacy preserving federated learning solution

v6-boilerplate-py

This algoithm is part of the vantage6 solution. Vantage6 allowes to execute computations on federated datasets. This repository provides a boilerplate for new algorithms.

Usage

First clone the repository.

# Clone this repository
git clone https://github.com/IKNL/v6-boilerplate-py

Rename the directories to something that fits your algorithm, we use the convention v6-{name}-{language}. Then you can edit the following files:

Dockerfile

Update the ARG PKG_NAME=... to the name of your algorithm (preferable the same as the directory name).

LICENCE

Determine which license suits your project.

{algorithm_name}/__init__.py

Contains all the methods that can be called at the nodes. All regular definitions in this file that have the prefix RPC_ are callable by an external party. If you define a master method, it should not contain the prefix! The master and regular definitions both have there own signature. Master definitions have a client and data argument (and possible some other arguments), while the regular definition only has the data argument. The data argument is a pandas dataframe and the client argument is a ClientContainerProtocol or ClientMockProtocol from the vantage6-toolkit. The master and regular definitions signatures should look like:

def some_master_name(client, data, *args, **kwargs):
    # do something
    pass

def RPC_some_regular_method(data, *args, **kwargs):
    # do something
    pass

setup.py

In order for the Docker image to find the methods the algorithm needs to be installable. Make sure the name matches the ARG PKG_NAME in the Dockerfile.

Read more

See the documentation for detailed instructions on how to install and use the server and nodes.

vantage6