[FINCN-353] Research and implement a federated learning model

• Type: Sub-task
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Assignee: Unassigned
• Reporter: Yemdjih Kaze Nasser
• Created: 13/Jun/22 01:03
• Updated: 13/Jun/22 01:07

Description

With the deficiency of consistent data and available datasets, it's really challenging to build a good model for our credit scorecard. This gives the advantage to use federated learning to train our models based on data in client devices without copying the data to a central unit hence preserving the client anonymously while building a robust model.

The aim here will be to research and implement a federated learning algorithm/model. For starters, this can be done with a simulation environment locally.

FYI: I think it will be easier to implement this using [TensorFlow Federated|https://www.tensorflow.org/federated.] Here is guide that can help to get started: [Federated Learning for Image Classification  |  TensorFlow Federated|https://www.tensorflow.org/federated/tutorials/federated_learning_for_image_classification] 

TensorFlow is distributed under the Apache 2.0 Licence so we shouldn't face any problem with Licencing conflicts.

Issue Links

• blocks: FINCN-354 Integrate and Deploy Federated model