Efficient Solutions for Machine Learning at the Edge

Explore innovative solutions for implementing machine learning on resource-constrained edge devices in this seminar by Prof. Saurav Prakash from IIT Madras. Discover how to overcome privacy concerns that restrict data sharing across multiple device owners while addressing the significant resource constraints and heterogeneity that limit edge devices' ability to handle large AI models. Learn about cutting-edge approaches to enable efficient and privacy-preserving machine learning in diverse edge computing environments. Examine the critical challenge of federated learning where each edge device can only train small local models while contributing to a larger global model. Understand the dynamic AI-powered data ecosystem created by the rapid growth of edge devices and its potential for societal advancement. Gain insights into information and coding theory applications, machine unlearning techniques, federated learning architectures, and hyperbolic geometry approaches that address these computational challenges. The presentation covers practical solutions for enabling distributed machine learning across heterogeneous edge networks while maintaining data privacy and computational efficiency.