Learn the principles of federated learning with an IBM python framework. This new machine learning paradigm means that training data does not need to be transmitted to a central place to avoid exposing private information and to comply with privacy regulations. Multiple parties are able to train a collaborative machine learning model by exchanging model updates with an aggregator.
Federated learning (FL) is a new machine learning paradigm where training data does not need to be transmitted to a central place to avoiding exposing private information and to comply with privacy regulations. In these settings, multiple parties are able to train a collaborative machine learning model by simply exchanging model updates with an aggregator. Because data that would be otherwise inaccessible is now available for training, FL enables the creation of more generalizable and accurate models. IBM Federated Learning is a Python framework created to bring that capability to all enterprises.
In this talk, IBM Research AI Security and Privacy Solutions Manager Nathalie Baracaldo will explain the principles of IBM Federated Learning and then show how data scientists can use this library in the Jupyter ecosystem to bring to unlock data previously inaccessible.