Federated Learning

What You'll Learn

• Understand components of federated learning systems and customize them for efficient model training.
• Utilize federated learning for LLMs to manage privacy and efficiency challenges effectively.
• Learn techniques like parameter-efficient fine-tuning and differential privacy to enhance security and efficiency.

About This Course

This two-part course covers the fundamentals and advanced applications of federated learning using the Flower framework. Federated learning enables training across distributed data sources, enhancing data privacy and security. In part one, you’ll set up and tune federated learning systems, and in part two, you’ll apply these methods to large language models (LLMs).

• Federated Training: Learn federated learning’s process and apply it to various models, including language, speech, and vision models.
• Private Enhancing Technologies (PETs): Integrate differential privacy into federated learning projects for enhanced data security.
• Data Bandwidth Optimization: Optimize bandwidth in federated learning by reducing update size and communication frequency.
• Federated Fine-tuning of LLMs: Discover techniques for fine-tuning LLMs with private data, enhancing data security and efficiency with federated approaches.

Course Outline

• Introduction
  Overview of federated learning and its applications in model training.
• Why Federated Learning
  Discussion of federated learning’s benefits, use cases, and practical applications.
• Federated Training Process
  Step-by-step guide to setting up and executing a federated training project.
• Tuning
  Techniques for customizing and optimizing federated learning systems.
• Data Privacy
  Introduction to differential privacy and privacy-preserving technologies in federated learning.
• Bandwidth
  Methods for optimizing bandwidth usage in federated learning.
• Conclusion
  Summary of key takeaways and next steps for implementing federated learning in real-world applications.

Who Should Join?

This course is designed for individuals with a background in Python and machine learning, an understanding of LLMs, and an interest in building models with federated learning using the Flower framework.