Apostolos Pyrgelis

Title: Apostolos Pyrgelis: Innovator in Privacy-Preserving Machine Learning

Introduction

Apostolos Pyrgelis is a notable inventor based in Bretigny-Sur-Morrens, Switzerland. He has made significant contributions to the field of machine learning, particularly in the area of privacy-preserving distributed training. With a total of 2 patents, Pyrgelis is at the forefront of innovations that enhance data security in neural network models.

Latest Patents

Pyrgelis's latest patents include a "System and method for privacy-preserving distributed training of neural network models on distributed datasets." This invention outlines a computer-implemented method and a distributed computer system designed for the secure training of a global neural network model. The system involves data providers that are communicatively coupled, each possessing a local training dataset and a vector of output labels. The method employs a cryptographic distributed secret key and a corresponding collective cryptographic public key, ensuring that the weights of the global model remain encrypted. Each data provider computes and aggregates encrypted local gradients, which are then used to update the global model.

Another significant patent is the "System and method for privacy-preserving distributed training of machine learning models on distributed datasets." This system features multiple data providers, each with a local model and training dataset. The models are encrypted using a collective public key, and each data provider trains its local model iteratively. The local models are then combined to update the current global model, which is shared with other data providers.

Career Highlights

Apostolos Pyrgelis is affiliated with the École Polytechnique Fédérale de Lausanne, a prestigious institution known for its cutting-edge research and innovation. His work focuses on developing systems that prioritize privacy while enabling effective machine learning processes.

Collaborations

Pyrgelis collaborates with esteemed colleagues such as David Froelicher and Juan Ramon Troncoso-Pastoriza. Their combined expertise contributes to advancing the field of privacy-preserving technologies in machine learning.

Conclusion

Apostolos Pyrgelis is a pioneering inventor whose work in privacy-preserving distributed training is shaping the future of machine learning. His contributions are vital for enhancing data security in neural networks, making significant strides in the field.