FEDEA

Description

The FEDEA project focuses on developing an advanced Federated Learning platform applied to natural language technologies, and specifically to voice systems, with the aim of enabling the collaborative training of AI models without the need to share personal data. Given the limitations of traditional Deep Learning—which requires the centralization of large volumes of data—FEDEA proposes an alternative that fully complies with the GDPR and guarantees privacy while maintaining the quality of the models. The project combines Federated Learning with Active Learning techniques to optimize training and validate the platform in real-world scenarios, facilitating its future industrialization and commercialization.

Objective

The main objective of the FEDEA project is to design, develop, and validate an innovative platform that enables decentralized and collaborative training of voice AI models through Federated Learning, ensuring data privacy and complying with GDPR obligations. The solution will integrate Active Learning techniques to improve training efficiency and will be evaluated in real-world environments to ensure its robustness, scalability, and industrial applicability.

Actions

• Analyze the limitations of traditional AI methods regarding GDPR compliance.
• Study Federated Learning technologies applied to voice data.
• Design the architecture of the federated training platform.
• Integrate Active Learning techniques to optimize the training process.
• Develop innovative features compared to existing federated solutions.
• Implement collaborative training in real distributed environments.