Federated learning (FL) is a machine learning approach for training models on decentralized edge devices without sharing raw data. Within this framework, organizations can collaborate on model development, models can gain experience from a vast range of data located at different sites, and data privacy can be preserved.

But FL implementation comes with challenges.

This session addresses one of them: optimizing FL workload performance on CPUs and GPUs.

It takes advantage of the implementation and results of a recent ASUS* FL solution for the healthcare industry that significantly boosts efficiency and performance by using Intel® Xeon® processors with Intel oneAPI and Intel® tools.

This session shows:

• An overview of the ASUS solution—the approach, methodologies, and results.
• How to use Intel® AI software stacks and tools to analyze and enhance the performance of convolutional neural network (CNN) models, such as VGG19 and EfficientNet, in real medical scenarios.
• How to review performance issues of AI framework and verify the compatibility of the desired hardware.
• How to collaboratively train a CNN model without sharing private data and images using the open source FL framework Flower.

Skill level: All
 

• Get Intel® oneAPI Deep Neural Network Library as a stand-alone product or as part of the Intel® oneAPI Base Toolkit.
• Get Intel® VTune™ Profiler as a stand-alone product or as part of the Intel oneAPI Base Toolkit.
• Intel® Distribution for Python*
• Intel® Extension for PyTorch*
   

• ASUS—Unleash the Power of Intel CPUs in Federated Learning: A Comprehensive Evaluation (PDF)