Praneeth Vepakomma

Title: Praneeth Vepakomma: Innovator in Federated Learning and Deep Learning Privacy

Introduction

Praneeth Vepakomma is an accomplished inventor based in Weymouth, MA (US). He has made significant contributions to the fields of federated learning and deep learning privacy, holding three patents that showcase his innovative approaches to complex problems in these areas.

Latest Patents

His latest patents include "Systems and methods for providing a modified loss function in federated-split learning." This patent discloses a method that involves training a part of a deep learning network at a client up to a split layer. The server then completes the training by forward propagating the output received at the split layer to the last layer. The server calculates a weighted loss function for the client and stores it. After all clients have their respective loss functions stored, the server averages these functions and back propagates gradients based on the average loss value. This innovative approach enhances the efficiency of federated learning systems.

Another notable patent is "Methods and apparatus for reducing leakage in distributed deep learning." This invention addresses the challenge of preventing attackers from reconstructing raw data from activation outputs of an intermediate layer. The method minimizes distance correlation between raw data and activation outputs, ensuring attribute-level privacy. Clients can calculate decorrelated representations of raw data before sharing information, enhancing security in distributed deep learning networks.

Career Highlights

Praneeth has worked with notable organizations such as Tripleblind, Inc. and the Massachusetts Institute of Technology. His experience in these institutions has allowed him to collaborate with leading experts in the field and contribute to groundbreaking research.

Collaborations

He has collaborated with talented individuals like Gharib Gharibi and Ravi Patel, further enriching his work and expanding the impact of his inventions.

Conclusion

Praneeth Vepakomma's innovative contributions to federated learning and deep learning privacy demonstrate his commitment to advancing technology in meaningful ways. His patents reflect a deep understanding of complex systems and a dedication to enhancing security and efficiency in data processing.