DLFi – The Distributed Learning Framework



Conventional ML training has the safety concern of collecting all data in a single place and violating the privacy of the data owners. DLFi aims to solve this. 

DLFi is a Privacy-Preserving AI-as-as-Service (PP-AIaaS) solution. It provides a training environment on remote sites without necessarily displacing or transferring the data. It enables data providers to train a ML model without revealing their business-critical data to each other. DLFi offers communication-efficiency and guarantees the privacy of data owners.

Features

  • Privacy-Preserving
  • Cloud-Native
  • Modular and Pluggable
  • Customized Visualization Dashboard
  • GPU-Acceleration Support

Applications

  • ML Model Training over Geo-distributed Data Sources
  • Collaborative ML Model Training in Multi-domain Multi-vendor and Disaggregated Networks
  • Shared Governance and Ownership of ML Models

Implemented Use-cases

  • QoT Estimation in Multi-domain Multi-vendor Optical Networks
  • Vision Inspection for Quality Assurance in Factory Shop Floors

Previous Demonstrations

  • Demonstration of Federated Learning over Edge-Computing Enabled Metro Optical Networks at ECOC 2020

Flyer