Abstract
Recommendation systems can play an important role in today’s digital content platforms by supporting the suggestion of relevant content in a personalised manner for each customer. Such content customisation has not been consistent across most media domains, and particularly on radio streaming and gaming aggregators, which are the two real-world application domains focused in this work. The challenges faced in these application areas are the dynamic nature of user preferences and the difficulty of generating recommendations for less popular content, due to the overwhelming choice and polarisation of available top content. We present the design and implementation of a Reinforcement Learning-based Recommendation System (RLRS) for web applications, using a Deep Deterministic Policy Gradient (DDPG) agent and, as a reward function, a weighted sum of the user Click Distribution (CD) across the recommended items and the Dwell Time (DT), a measure of the time users spend interacting with those items. Our system has been deployed in real production scenarios with preliminary but promising results. Several metrics are used to track the effectiveness of our approach, such as content coverage, category diversity, and intra-list similarity. In both scenarios tested, the system shows consistent improvement and adaptability over time, reinforcing its applicability. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

Abstract
All types of sports are potential application scenarios for automatic and real-time visual object and event detection. In rink hockey, the popular roller quad skate variant of hockey team sports, it is of great interest to automatically track player’s movements and positions, player’s sticks and, also, making other judgments, such as being able to locate the ball. In this work, we introduce a real-time pipeline composed by an object detection model, created specifically for rink hockey games, followed by a knowledge-based event detection module. Even in the presence of occlusions and quick motions, our deep learning object detection model effectively identifies and tracks, in real-time, important visual elements such as: ball; players; sticks; referees; crowd; goalkeeper; and goal. Using a curated dataset composed by a collection of videos of rink hockey, comprising 2525 annotated frames, we trained and evaluated the algorithm performance and compare it to state of the art object detection techniques. Our object detection model, based on YOLOv7, presents a global accuracy of 80%, and presents a good performance in terms of accuracy and speed, according to our results, making it a good choice for rink hockey applications. In our initial tests, the event detection module successfully detected one important event type in rink hockey games, the occurrence of penalties.

Abstract
This study investigates the effectiveness of Transformer-based models for retail demand forecasting. We evaluated vanilla Transformer, Informer, Autoformer, PatchTST, and temporal fusion Transformer (TFT) against traditional baselines like AutoARIMA and AutoETS. Model performance was assessed using mean absolute scaled error (MASE) and weighted quantile loss (WQL). The M5 competition dataset, comprising 30,490 time series from 10 stores, served as the evaluation benchmark. The results demonstrate that Transformer-based models significantly outperform traditional baselines, with Transformer, Informer, and TFT leading the performance metrics. These models achieved MASE improvements of 26% to 29% and WQL reductions of up to 34% compared to the seasonal Na & iuml;ve method, particularly excelling in short-term forecasts. While Autoformer and PatchTST also surpassed traditional methods, their performance was slightly lower, indicating the potential for further tuning. Additionally, this study highlights a trade-off between model complexity and computational efficiency, with Transformer models, though computationally intensive, offering superior forecasting accuracy compared to the significantly slower traditional models like AutoARIMA. These findings underscore the potential of Transformer-based approaches for enhancing retail demand forecasting, provided the computational demands are managed effectively.

Abstract
Retailers depend on accurate sales forecasts to effectively plan operations and manage supply chains. These forecasts are needed across various levels of aggregation, making hierarchical forecasting methods essential for the retail industry. As competition intensifies, the use of promotions has become a widespread strategy, significantly impacting consumer purchasing behavior. This study seeks to improve forecast accuracy by incorporating promotional data into hierarchical forecasting models. Using a sales dataset from a major Portuguese retailer, base forecasts are generated for different hierarchical levels using ARIMA models and Multi-Layer Perceptron (MLP) neural networks. Reconciliation methods including bottom-up, top-down, and optimal reconciliation with OLS and WLS (struct) estimators are employed. The results show that MLPs outperform ARIMA models for forecast horizons longer than one day. While the addition of regressors enhances ARIMA's accuracy, it does not yield similar improvements for MLP. MLPs present a compelling balance of simplicity and efficiency, outperforming ARIMA in flexibility while offering faster training times and lower computational demands compared to more complex deep learning models, making them highly suitable for practical retail forecasting applications.

Abstract
Purpose 2-[F-18]FDG PET/CT plays an important role in the management of pulmonary nodules. Convolutional neural networks (CNNs) automatically learn features from images and have the potential to improve the discrimination between malignant and benign pulmonary nodules. The purpose of this study was to develop and validate a CNN model for classification of pulmonary nodules from 2-[F-18]FDG PET images.Methods One hundred thirteen participants were retrospectively selected. One nodule per participant. The 2-[F-18]FDG PET images were preprocessed and annotated with the reference standard. The deep learning experiment entailed random data splitting in five sets. A test set was held out for evaluation of the final model. Four-fold cross-validation was performed from the remaining sets for training and evaluating a set of candidate models and for selecting the final model. Models of three types of 3D CNNs architectures were trained from random weight initialization (Stacked 3D CNN, VGG-like and Inception-v2-like models) both in original and augmented datasets. Transfer learning, from ImageNet with ResNet-50, was also used.Results The final model (Stacked 3D CNN model) obtained an area under the ROC curve of 0.8385 (95% CI: 0.6455-1.0000) in the test set. The model had a sensibility of 80.00%, a specificity of 69.23% and an accuracy of 73.91%, in the test set, for an optimised decision threshold that assigns a higher cost to false negatives.Conclusion A 3D CNN model was effective at distinguishing benign from malignant pulmonary nodules in 2-[F-18]FDG PET images.

Abstract
After the learning process, certain types of images may not be modeled correctly because they were not well represented in the training set. These failures can then be compensated for by collecting more images from the real-world and incorporating them into the learning process - an expensive process known as active learning. The proposed twist, called active supervision, uses the model itself to change the existing images in the direction where the boundary is less defined and requests feedback from the user on how the new image should be labeled. Experiments in the context of class imbalance show the technique is able to increase model performance in rare classes. Active human supervision helps provide crucial information to the model during training that the training set lacks.