The PROTEAS EDF Project is a landmark initiative for special operations in areas such as EfficientNet, XGBoost, GPU, and CPU technologies.

The PROTEAS Project achieved a significant research milestone, highlighting the cutting-edge work of its partner, UBITECH which prepared the scientific paper titled as “On Energy-aware and Verifiable Benchmarking of Big Data Processing targeting AI Pipelines”. This breakthrough scientific paper was presented at the IEEE Big Data 2024 conference, held on December 15, 2024, in Washington D.C., highlighting a pioneering framework for evaluating hardware performance, energy efficiency, and AI model effectiveness on both tabular and image datasets. The authors of the paper were George Theodorou, Sophia Karagiorgou and Christos Kotronis.

Specifically, a new framework was presented for comparing hardware performance, energy consumption, and the efficiency of AI models on tabular and image data. The key findings of the study identified EfficientNet and XGBoost as top-performing models, while in terms of energy consumption, GPUs outperformed in image processing tasks, and CPUs were more efficient in tabular data analysis.

The proposed framework enables real-time monitoring of energy usage and computational resources in containerized environments, facilitating scalable and efficient AI deployment, especially for edge devices with limited resources. By integrating information theory and probability models, the framework enhances understanding of AI model behavior in edge-to-cloud applications.

This outcome advances sustainable and efficient Big Data AI workflows for real-world scenarios, proposing essential models and energy consumption patterns for future applications.

The Scientific Paper is published in the proceedings of IEEE Big Data 2024 and is also accessible through:  https://ieeexplore.ieee.org/abstract/document/10826014