Abstract
The paper presents the comparison of accuracy in the Speech Emotion Recognition task using the Hamming and Hanning windows for framing the speech and determining the spectrogram to be used as input of a convolutional neural network. The detection of between 4 and 10 emotional states was tested for both windows. The results show significant differences in accuracy between the two window types and provide valuable insights for the development of more efficient emotional state detection systems. The best accuracy between 4 and 10 emotions was 64.1% (4 emotions), 57.8% (5 emotions), 59.8% (6 emotions), 48.4% (7 emotions), 47.8% (8 emotions), 51.4% (9 emotions), and 45.9% (10 emotions). These accuracy is at the state-of-the art level.

Abstract
The increasing complexity of integrated energy systems with the electric power grid requires innovative control solutions for efficient management of smart buildings and distributed energy resources. Accurately predicting weather conditions and electricity demand is crucial to make such informed decisions. Machine learning has emerged as a powerful solution to enhance prediction accuracy by harnessing advanced algorithms, but often requires complex parameterizations and ongoing model updates. The Lawrence Berkeley National Laboratory's Autonomous Forecast Framework (AFF) was developed to greatly simplify this process, providing reliable and accurate forecasts with minimal user interaction, by automatically selecting the best model out of a library of candidate models. This work expands on the AFF by not only selecting the best model, but assembling a blend of multiple models into a hybrid forecast model. The validation within this work has shown that this combination of models outperformed the selected best model of the AFF 31%, while providing greater resilience to individual model's forecast error.

Abstract
The paper presents a modification of the Magnitude Optimum Multiple Integration (MOMI) method process non-parametric data in the frequency domain instead of the time domain The required frequency data are obtained directly from the filtered amplitude -shifted process step response and have been shown to be relatively insensitive to normally distributed process noise. All calculations, including the calculation of the PID controller parameters, are performed analytically. The closed loop responses to tested processes with added normally distributed noise were relatively fast with small or no overshoot, all according to the Magnitude Optimum (MO) method. The proposed method is not limited to open loop step responses or to the PID controller structure.

Abstract
Integrating processes can be found in various industries. The main characteristic of such processes is that a limited process input can cause an unlimited process output. In general, they are more difficult to control compared to stable processes. The recently developed Magnitude optimum multiple integration tuning method for integrating processes provides very good closed -loop responses. However, it uses a reference -weighting 2-DOF PI(D) controller structure where the weighting parameters for the P and D term of the controller are equal (therefore the user can only change one parameter). Another drawback of the existing method is that it needs to find the roots of the fourth -order algebraic equation. The method proposed here does not require finding these roots and provides better tracking compared to the original method while maintaining optimal disturbance rejection for different integrating process models.

Abstract
Recently introduced Generalized Pre-trained Transformers (GPT) and conversional chatbots such as ChatGPT are causing deep society transformations. The incorporation of these Artificial Intelligence technologies can be beneficial in multiple science and development areas including Control Engineering. The evaluation of GPTs within Control Engineering Education and PID control is addressed in this work. Different types of interactions with GPTs are evaluated and the use of a personalized GPT for PID tuning explored. Copyright (C) 2024 The Authors. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0/)

Abstract
With the serious danger to nature and humanity that forest fires are, taken into consideration, this work aims to develop an artificial intelligence model capable of accurately predicting the forest fire risk in a certain region based on four different factors: temperature, wind speed, rain and humidity. Thus, three models were created using three different approaches: Artificial Neural Networks (ANN), Random Forest (RF), and K-Nearest Neighbor (KNN), and making use of an Algerian forest fire dataset. The ANN and RF both achieved high accuracy results of 97%, while the KNN achieved a slightly lower average of 91%. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.