Hybrid L‑BFGS + Adam Optimization for Speech Emotion Recognition
During the ICSEng 2025 conference, NASK PhD student Paweł Czyż presented his research on the processes of emotion recognition in speech using AI.
On October 27–28, 2025, the International Conference on Systems Engineering (ICSEng 2025) was held in Warsaw, recognized as a leading international forum for systems engineering, attracting scientists and experts from around the world. The conference was organized by experienced specialists in systems engineering and artificial intelligence, ensuring high scientific quality and prestige.
The oral presentation was delivered by Paweł Czyż, presenting research on a hybrid optimization approach combining L‑BFGS and Adam for speech emotion recognition. The work has been accepted for publication in the prestigious Springer Lecture Notes in Networks and Systems series, highlighting its scientific contribution.
The study aimed to combine the advantages of the adaptive Adam optimizer with quasi‑Newton second-order L‑BFGS to improve the stability and speed of neural network training for analyzing speech signals. Experiments demonstrated that the hybrid algorithm outperforms the classical approach, achieving faster convergence and lower loss values.
The research results may find applications in emotionally aware systems and human‑computer interfaces. The presentation attracted considerable interest from participants in the fields of systems engineering, artificial intelligence, and optimization.
