Self-Paced-Training Tutorials
Federated Learning with NVIDIA FLARE: Notebooks and Videos
Welcome to the five-part course on Federated Learning with NVIDIA FLARE! This course covers everything from the fundamentals to advanced applications, system deployment, privacy, security, and real-world industry use cases. It has 100+ notebooks and 80 videos.
Note
These notebooks were developed with NVFlare 2.6. Not all content reflects the latest APIs; some examples may need adjustments for newer versions.
What You’ll Learn
Fundamentals: Understand federated learning and decentralized training concepts.
System Architecture: Learn about NVIDIA FLARE system architecture, deployment, and user interactions.
Privacy & Security: Explore privacy and security challenges, solutions, and enterprise-grade protections.
Advanced Topics: Dive into algorithms (FedOpt, FedProx, etc.), workflows (cyclic, split, swarm), LLM training, and XGBoost.
Industry Applications: Discover real-world use in healthcare, life sciences, and finance.
Practical Skills: Transition from standard ML code to federated workflows; customize client/server logic, job structure, and configuration.
Comprehensive Resources: Access over 100 notebooks and 88 videos for a thorough learning experience.
Tip
While each notebook is self-contained and can be run independently, for best results, follow the sequence to build a strong foundation.
Course Outline
Part 1: Introduction to Federated Learning
This section provides a hands-on introduction to federated learning using NVIDIA FLARE. You will learn how to run and develop federated learning applications, with practical examples and clear guidance for both beginners and experienced practitioners.
What You’ll Learn
The basics of federated learning and its advantages
How to use NVIDIA FLARE to train and deploy federated learning models
Transitioning from standard ML code to federated learning workflows
Customizing client and server logic in NVIDIA FLARE
Understanding job structure, configuration, and statistics
Chapter 1: Running Federated Learning Applications
Train an image classification model with PyTorch using NVIDIA FLARE
Convert standard PyTorch training code to federated learning code
Customize client and server logic in NVIDIA FLARE
Explore job structure and configuration for federated learning
Chapter 2: Developing Federated Learning Applications
Perform federated statistics for both image and tabular data
Convert PyTorch Lightning and traditional ML code to federated learning workflows with NVIDIA FLARE
Use the NVIDIA FLARE Client API for advanced customization
Part 2: Federated Learning System
This section explores the architecture and deployment of federated computing systems using NVIDIA FLARE. You will gain practical knowledge on system setup, user interaction, and monitoring tools for federated learning environments.
What You’ll Learn
NVIDIA FLARE system architecture and core concepts
How to set up and simulate a federated computing system (local deployment)
User interaction methods: admin console, Python API, and CLI
Monitoring system events with Prometheus and Grafana
Chapter 3: Federated Computing Platform
Understand the NVIDIA FLARE federated computing platform and its components
Learn about system roles, communication, and workflow
Chapter 4: Setup Federated Computing System
Step-by-step guide to setting up a federated computing system with NVIDIA FLARE
Simulate deployments and interact with the system using various tools
Part 3: Security and Privacy
Federated learning enables decentralized model training while preserving data privacy, making it ideal for sensitive domains like healthcare and finance. However, federated learning introduces security and privacy risks, such as data leakage, adversarial attacks, and model integrity threats.
What You’ll Learn
Privacy risks and attack vectors in federated learning
Protections: differential privacy, secure aggregation, homomorphic encryption
Security challenges: adversarial attacks, unauthorized access, communication threats
Security solutions: authentication, RBAC, encrypted communication, trust mechanisms
How NVIDIA FLARE implements robust security and privacy for federated learning
Chapter 5: Privacy in Federated Learning
Understand privacy risks and attacks in federated learning
Explore privacy-preserving techniques with NVIDIA FLARE
Chapter 6: Security in Federated Computing System
Learn about security threats and solutions in federated learning
See how NVIDIA FLARE enforces secure communication, authentication, and access control
Part 4: Advanced Topics in Federated Learning
This section explores advanced topics and techniques in federated learning using NVIDIA FLARE. You will learn about cutting-edge algorithms, workflows, large language model (LLM) training, secure XGBoost, and the distinction between high-level and low-level APIs.
What You’ll Learn
Advanced federated learning algorithms: FedOpt, FedProx, and more
Workflows: cyclic, split learning, swarm learning
Training and fine-tuning large language models (LLMs) with NVIDIA FLARE
Secure federated XGBoost
High-level vs. low-level APIs in NVIDIA FLARE
Chapter 7: Federated Learning Algorithms and Workflows
Explore various federated learning algorithms and workflow strategies with NVIDIA FLARE
Chapter 8: Federated LLM Training
Learn how to train and fine-tune large language models in a federated setting with NVIDIA FLARE
Chapter 9: NVIDIA FLARE Low-level APIs
Discover the power and flexibility of NVIDIA FLARE’s low-level APIs
Chapter 10: Federated XGBoost
Step-by-step guide to secure federated XGBoost with NVIDIA FLARE
Part 5: Federated Learning Applications in Industries
This section demonstrates how NVIDIA FLARE is applied in real-world industry settings, focusing on healthcare, life sciences, and financial services. Learn how federated learning enables collaboration, privacy, and innovation across organizations.
What You’ll Learn
How NVIDIA FLARE powers collaborative machine learning in healthcare and life sciences, including:
Medical image analysis (e.g., cancer detection, radiology)
Survival analysis (e.g., Kaplan-Meier)
Genomics and multi-institutional research
Drug discovery
Financial services applications, such as:
Fraud detection
Anomaly detection in transactions
Chapter 11: Federated Learning in Healthcare and Life Sciences
Use cases for NVIDIA FLARE in medical research, diagnostics, and drug discovery
How to train robust, privacy-preserving models across hospitals and research centers
Chapter 12: Federated Learning in Financial Services
Collaborative model training for fraud detection, credit risk, and regulatory compliance
Getting Started
Start with any part or topic of interest, or follow the sequence for a comprehensive journey.
Refer to the official NVIDIA FLARE documentation for deeper dives and troubleshooting.
Happy learning!