New Scientific Publication on Bayesian Optimization for Efficient Machine Learning Model

UltraSense partner University of Oulu has just published a scientific article on Bayesian optimization introducing a novel Beta Product Kernel designed for bounded domains.

As machine learning models become increasingly complex, their deployment in real‑world applications—especially in constrained environments—poses significant challenges. Large models often require substantial computational power, memory and energy resources, limiting their practical usability beyond laboratory settings.

The proposed methodology addresses this challenge by introducing a novel optimization approach that improves how machine learning models can be efficiently configured and applied in constrained settings. The method has been evaluated across different scenarios, including applications related to model compression, showing its potential to support more efficient and scalable solutions. By enabling improved performance while supporting efficiency, the approach contributes to the development, deployment and real‑world exploitation of machine learning models, making them more suitable for practical and scalable applications.

This research contributes directly to UltraSense objectives by supporting the creation of lighter, more efficient and deployable AI solutions, an essential requirement for sensing, data processing and intelligent systems operating under real‑world constraints.

The paper is available as a preprint on arXiv and represents an important step towards bridging the gap between high‑performance machine learning research and its practical implementation.

📄 Find out more!
👉 https://arxiv.org/pdf/2506.16316