Iskanje po katalogu digitalne knjižnice

Opis: Nowadays, technological progress has promoted the integration of artificial intelligence into modern human lives rapidly. On the other hand, extreme weather events in recent years have started to influence human well-being. As a result, these events have been addressed by artificial intelligence methods more and more frequently. In line with this, the paper focuses on searching for predicting the air temperature in a particular Slovenian micro-location by using a weather prediction model Maximus based on a longshort term memory neural network learned by the long-term, lower-resolution dataset CERRA. During this huge experimental study, the Maximus prediction model was tested with the ICON-D2 general-purpose weather prediction model and validated with real data from the mobile weather station positioned at a specific micro-location. The weather station employs Internet of Things sensors for measuring temperature, humidity, wind speed and direction, and rain, while it is powered by solar cells. The results of comparing the Maximus proposed prediction model for predicting the air temperature in micro-locations with the general-purpose weather prediction model ICON-D2 has encouraged the authors to continue searching for an air temperature prediction model at the micro-location in the future.
Ključne besede: long short-term memory neural networks, air temperature, micro-location, prediction, weather, Internet of Things
Objavljeno v DKUM: 25.09.2025; Ogledov: 0; Prenosov: 10
Celotno besedilo (8,81 MB)

Opis: Spatio-temporal forecasting is a rapidly evolving field, accelerated by the increasing accessibility of sensoring infrastructure and computational hardware, capable of processing the large amount of sampled data. Applications of spatio-temporal forecasts range from traffic, weather, air pollution forecasting and others. Emerging technologies employ deep learning architectures, such as graph, convolutional, recurrent and transformer neural networks. While the state-of-the-art methods provide accurate time series predictions, they are typically limited to providing forecasts only for the direct locations of sampling, whereas coverage of the entire area is often desired by the applications. In this work, we propose a method that addresses this challenge and improves on the shortcomings of related works, which have already tackled the task. The proposed graph convolutional recurrent neural network based method provides forecasts for arbitrary geolocations without available measurement data, formulating predictions based on contextual information of target geolocations and the time series data of nearby measurement geolocations. We evaluate the method on three real-world datasets from meteorological, traffic and air pollution domains, and gauge its performance against the state-of-the-art spatio-temporal forecasting methods. The proposed method achieves 12.26 %, 66.97 % and 42.89 % improvements in the mean absolute percentage errors on the three aforementioned datasets, compared to the best performing state-of-the-art method GConvGRU.
Ključne besede: spatio-temporal forecasting, graph recurrent neural networks, sparse geospatial sensor networks
Objavljeno v DKUM: 25.07.2025; Ogledov: 0; Prenosov: 2
Celotno besedilo (5,19 MB)

Opis: Machine learning (ML) has demonstrated remarkable ability to uncover hidden patterns in data. However, the presence of biases and discrimination originating from the data itself and, consequently, emerging in the ML outcomes, remains a pressing concern. With the exponential growth of unstructured data, such as images, fairness has become increasingly critical, as neural network (NN) models may inadvertently learn and perpetuate societal and historical biases. To address this challenge, we propose a fairness-aware loss function that iteratively prioritizes the worst-performing sensitive group during NN training. This approach aims to balance treatment quality across sensitive groups, achieving fairer image classification outcomes while incurring only a slight compromise in overall performance. Our method, evaluated on the FairFace dataset, demonstrates significant improvements in fairness metrics while maintaining comparable overall quality. These trade-offs highlight that the minor decrease in overall quality is justified by the improvement in fairness of the models.
Ključne besede: fairness, search-basimage classification, machine learning, supervised learnign, neural networks
Objavljeno v DKUM: 23.04.2025; Ogledov: 0; Prenosov: 6
Celotno besedilo (2,01 MB)

Opis: Automated detection of ovarian follicles in ultrasound images is much appreciated when its effectiveness is comparable with the experts’ annotations. Today’s best methods estimate follicles notably worse than the experts. This paper describes the development of two-stage deeply-supervised 3D Convolutional Neural Networks (CNN) based on the established U-Net. Either the entire U-Net or specific parts of the U-Net decoder were replicated in order to integrate the prior knowledge into the detection. Methods were trained end-to-end by follicle detection, while transfer learning was employed for ovary detection. The USOVA3D database of annotated ultrasound volumes, with its verification protocol, was used to verify the effectiveness. In follicle detection, the proposed methods estimate follicles up to 2.9% more accurately than the compared methods. With our two-stage CNNs trained by transfer learning, the effectiveness of ovary detection surpasses the up-to-date automated detection methods by about 7.6%. The obtained results demonstrated that our methods estimate follicles only slightly worse than the experts, while the ovaries are detected almost as accurately as by the experts. Statistical analysis of 50 repetitions of CNN model training proved that the training is stable, and that the effectiveness improvements are not only due to random initialisation. Our deeply-supervised 3D CNNs can be adapted easily to other problem domains.
Ključne besede: 3D deep neural networks, 3D ultrasound images of ovaries, deep supervision, detection of follicles and ovaries, U-Net based architecture
Objavljeno v DKUM: 27.03.2025; Ogledov: 0; Prenosov: 11
Celotno besedilo (1,28 MB)
Gradivo ima več datotek! Več...

Opis: Bridges are typically designed by reputable, specialized engineering and design companies with years of experience. In these firms, experienced engineers share and pass on their knowledge to younger colleagues. However, when these experts retire, some of the knowledge is lost forever. As a subset of artificial intelligence methods, artificial neural networks (ANNs) can solve the problem of acquiring, transferring, and preserving specialized expert knowledge. This article describes the possible application of CAE ANN to acquire knowledge and to assist in the design of incrementally launched prestressed concrete bridges. Therefore, multidimensional graphs in the form of iso-curves of equal values were created, allowing practicing engineers to understand complex relationships between design parameters. The graphs also contain information about the reliability of the results, which is defined by an estimated parameter. The general rule is that results based on a larger number of actual data points are more reliable. Finally, an ANN BD assistant is proposed as an application that assists engineers and designers in the early stages of design and/or established engineers and designers in variant studies and design parameter optimization.
Ključne besede: artificial neural networks, bridge design, incremental launching method, expert knowledge, reliability of predictions, prestressed concrete bridges
Objavljeno v DKUM: 10.03.2025; Ogledov: 0; Prenosov: 18
Celotno besedilo (5,54 MB)
Gradivo ima več datotek! Več...