Machine learning applications in reservoir modeling are hindered by the limited availability of well log data, a common challenge in the oil and gas industry. We propose VAEc-tMC, a domain-informed Conditional Variational Autoencoder that generates synthetic well-log data conditioned on rock type. Addressing a critical gap by existing generative models that rely solely on statistical reconstruction, our model embeds geological domain knowledge into the latent space, and optimizes a modified objective with an adaptive Student-t reconstruction loss and a beta-weighted KL regularizer, improving stability under heavy-tailed data. When used for data augmentation, the synthetic samples preserve inter-log dependencies and substantially enhance downstream classification, accuracy 39→63%, F1-score 36→68%, AUC 0.46→0.80 on a held-out well. Beyond the geological context, the proposed approach illustrates a generalizable strategy where domain-aware generative models with adaptive loss functions provide a robust solution for data-efficient learning in scientific domains facing data scarcity, noise, and heavy-tailed distributions.

Compositional data consist of components expressed as proportions of a whole and carry only relative information. In statistical and machine learning contexts, these data require specialized handling due to their constant-sum constraint and non-Euclidean geometry. A common approach is the application of log-ratio transformations-such as the centered log-ratio (CLR)-to project compositional vectors into Euclidean space. While using CLR for inputs is well established, applying this transformation to compositional outputs remains underexplored. This study evaluates the predictive impact of using CLR on target variables in supervised learning, with a geochemical dataset containing lithogeochemical targets and physical-log predictors. Three deep learning architectures are assessed: CNNBiLSTM, SAIDNN, and MHA-BiRNN, each trained on raw and CLR-transformed outputs. Results consistently show that models trained directly on raw compositions outperform their CLR-transformed counterparts. We identify two causes: (i) the inverse CLR transformation redistributes prediction errors across components, reducing local precision, and (ii) zeros in compositional data introduce artifacts when preprocessed for log-ratio transformations. Furthermore, the models demonstrate good generalization on blind tests, especially when trained on raw data. These findings suggest that modern deep learning models can effectively learn from compositional outputs in their native form, avoiding distortions from transformation pipelines.

The generation of synthetic well log data is crucial for enhancing the understanding and exploration of subsurface reservoirs. This paper introduces a novel Generative Adversarial Network (GAN) model that incorporates a Principal Component Analysis (PCA)-based loss function to improve the quality of synthetic well log data. The proposed method is distinguished by its ability to generate complete well log datasets, rather than just individual logs or completing partial logs. Traditional GANs utilize cross-entropy loss but often fail to capture the complex structural patterns inherent in well logs. By integrating PCA into the loss function, our model not only distinguishes real from synthetic data but also ensures that the synthetic data retains the intrinsic variability and relationships observed in real logs. We validate our approach using histograms, correlation heatmaps, dimensionality reduction techniques (PCA and t-SNE), and a discriminative task. Results show that synthetic data generated with PCA-based loss aligns closely with real data, demonstrating superior preservation of statistical and structural characteristics. This advancement in synthetic data generation holds promise for enriching subsurface data analysis and exploration.

Nowadays, one of the most relevant challenges of a data center is to keep its information secure. To avoid data leaks and other security problems, data centers have to manage vulnerabilities, including determining the higher-risk vulnerabilities to prioritize. However, the current literature is scarce in the proposal of intelligent methods for the complex problem of vulnerabilities prioritization. Depending on the adopted metrics, the priority could shift, compromising simple sorting-based approaches and impairing the utilization of conflicting risk assessment metrics. Unlike the related work, this study proposes a multi-objective method that uses user-chosen vulnerabilities assessment metrics to output a complete list of these vulnerabilities ranked by their risk and overall impact in the context of an organization. The method includes a multi-objective large-scale optimization problem representation, a novel population initialization scheme, an expressive fitness function, a post-optimization process, and a custom way to select the best solution among the non-dominated ones. The dataset used in the experiments contains anonymized real-world information about database vulnerabilities obtained from a private organization. The experiments' results indicated that the proposed method can reduce the number of vulnerabilities needed to reach an organization's predefined security targets compared to the baselines simulating a security team's analysis. Multi-objective optimization achieved on average a 48,17% reduction in the vulnerabilities needed to reach the organization's target values compared to the baselines.

The continuous use of satellite networks in the Low Earth Orbit (LEO) has accumulated a large amount of space debris. Given the actual state of the orbit, these debris are a threat to the active systems and to the feasibility of future operations in LEO. Now, Active Debris Removal (ADR) missions must be conducted to mitigate the debris through forced deorbitation. The best documented approaches for the ADR mission planning made use of metaheuristics, modeling the ADR as a complex variant of the TSP. However, these approaches usually fail to deal some of the ADR problem dynamics, such as large instances, mission constraints or the debris motion. In this paper we propose heuristic of continuous improvement on a genetic-based solution. Our work advances the state of the art by dealing with large real world instances, modeling all the constraints and considering the problem time dependence (motion). Experiments were conducted to evidence the improvements over the literature. With the ability of generating time-dependent results for scenarios with thousands of debris in a feasible time, our approach yielded missions 96.33 % more effective at the cleaning job than the present ones on the literature.

Multi-objective decision-making and dynamic short-est paths are two areas of research widely studied and of great importance for computer science, engineering, and economics. Both areas have seen a remarkable evolution in their algorithms in the last decades and have contributed to several applications in real life. The applications range from cost reduction in the expansion of telecommunications and electrical networks to the accessibility of people and autonomous vehicles. However, despite their importance, the investigation of methods at the intersection of these two areas has not been explored in the literature. Problems at this intersection are characterized by graphs whose topology can change over time (i.e., edges can be inserted or deleted online) and whose edges' costs are defined by more than a single criterion or objective. In this paper, we introduce the multi- objective dynamic shortest path problem (MODSP) and present the first algorithm to solve it. In particular, we formally define the MODSP problem and explain its relation to multi-objective decision-making and dynamic shortest paths. Concerning the algorithm, we present the first single-source MODSP (SMDSP) approach as a first effort towards solving MODSP problems while avoiding the recomputation of the paths from scratch when the graph is updated. Finally, we perform an experimental evaluation of the SMDSP and a comparison with the state-of-art algorithm for multi-objective shortest path problems. The results of our experimental evaluation prove that in an environment subjected to constant updates, the SMDSP algorithm is more efficient.

Cardiovascular disease (CVD) is one of the main causes of death in the world. Coronary artery disease (CAD) is one of the CVD most common disorders. CAD is mainly caused by restrictions in the heart muscle blood flow supply, called atherosclerosis. The assessment of atherosclerosis is challenging and is currently primarily achieved with angiography, which is the gold standard for geometrical assessment. The angiography is completely dependent on the physician's lesion identification and visual assessment. In this paper, we tackle this problem by proposing an architecture for automatic lesion detection, called DeepCADD. DeepCADD proposes the use of instance segmentation, replaces an original Mask R-CNN's backbone, and trains the classification layer with an angiography dataset. The backbone was replaced by a ResNet-50 pre-trained with a dataset with 2,000 images of coronary artery segments. Powered by the skip connections from ResNet-50 network which propagates small image features to deeper layers. DeepCADD is comparable with the gold standard and we discussed the performance with similar studies. Moreover, we performed a validation with specialists to understand the architecture's real performance. Our results show that DeepCADD presents a good performance in the lesion identification, reducing the false negatives and reaching a sensitivity of approximately 0.89. These pieces of evidence suggest that DeepCADD can be used as a screening tool, contributing to the narrowed lesions identification, playing a crucial role in the angiography protocol automation.

With fast increasing urbanization levels, adaptive traffic signal control methods have great potential for optimizing traffic jams. In particular, deep reinforcement learning (RL) approaches have been shown to be able to outperform classic control methods. However, deep RL algorithms are often employed as black boxes, which limits their use in the real-world as the decisions made by the agents can not be properly explained. In this paper, we compare different function approximations methods used to estimate de action-value function of RL-based traffic controllers. In particular, we compare (i) their expressiveness, based on the resulting performance of the learned policies, and (ii) their explainability capabilities. To explain the decisions of each method, we use Shapley Additive Explanations (SHAP) to show the impact of the agent's state features on each possible action. This allows us to explain the learned policies with a single image, enabling an understanding of how the agent behaves in the face of different traffic conditions. In addition, we discuss the application of post-hoc explainability models in the context of adaptive traffic signal control, noting their potential and pointing out some of their limitations. Comparing our resulting methods to state-of-the-art adaptive traffic signal controllers, we saw significant improvements in travel time, speed score, and throughput in two different scenarios based on real traffic data.

Mental disorders affect a large number of people worldwide. In response to the increasing number of people affected by such illnesses, there has been an increased interest in the use of state-of-the-art technologies to mitigate its effects. This paper presents a Sequence Model for Stress Classification (MoStress), which is a novel pipeline for pre-processing physio-logical data collected from wearable devices and for identifying stress sequences using a recurrent neural network (RNN). Using the WESAD dataset, the RNN model achieved accuracy of 86% in a three-class classification problem (baseline vs. stress vs. amusement). When only considering the presence of stress or not, we achieved an accuracy of 96.5% as well as precision, recall, and f'1-score of 96%, 93%, and 94%, respectively. Those results are close to other papers using the same dataset, however, the neural network used on MoStress, is considerable simpler.

Abstract Real-world sequential decision-making tasks are generally complex, requiring trade-offs between multiple, often conflicting, objectives. Despite this, the majority of research in reinforcement learning and decision-theoretic planning either assumes only a single objective, or that multiple objectives can be adequately handled via a simple linear combination. Such approaches may oversimplify the underlying problem and hence produce suboptimal results. This paper serves as a guide to the application of multi-objective methods to difficult problems, and is aimed at researchers who are already familiar with single-objective reinforcement learning and planning methods who wish to adopt a multi-objective perspective on their research, as well as practitioners who encounter multi-objective decision problems in practice. It identifies the factors that may influence the nature of the desired solution, and illustrates by example how these influence the design of multi-objective decision-making systems for complex problems.

Abstract The recent paper “Reward is Enough” by Silver, Singh, Precup and Sutton posits that the concept of reward maximisation is sufficient to underpin all intelligence, both natural and artificial, and provides a suitable basis for the creation of artificial general intelligence. We contest the underlying assumption of Silver et al. that such reward can be scalar-valued. In this paper we explain why scalar rewards are insufficient to account for some aspects of both biological and computational intelligence, and argue in favour of explicitly multi-objective models of reward maximisation. Furthermore, we contend that even if scalar reward functions can trigger intelligent behaviour in specific cases, this type of reward is insufficient for the development of human-aligned artificial general intelligence due to unacceptable risks of unsafe or unethical behaviour.

The traffic assignment problem (TAP) consists of assigning routes to road users in order to minimize traffic congestion. Traditional methods for solving the TAP assume the existence of a central authority who computes and dictates routes to road users. Multi-agent reinforcement learning (MARL) approaches are more realistic in solving this kind of problem because they consider that road users (agents) have complete autonomy for choosing routes. However, MARL approaches usually require a long training period in order to compute the optimal routes, which could be a major limitation in more realistic traffic scenarios. In this paper, we tackle this problem by evaluating the performance of three conceptually different reward functions, namely: expert-designed rewards, difference rewards, and intrinsically motivated rewards. In particular, our focus lies on providing a deeper understanding of the impact of these reward functions on the agents' performance, thus contributing towards reducing congestion levels. To this end, we perform an extensive experimental evaluation on different road networks, including up to 360,600 concurrently learning agents. Our results show that, although the adopted reward functions were not able to speed up the learning process, the correct reward function choice plays an important role in the quality of the learned solution.

The idea of experience sharing between cooperative agents naturally emerges from our understanding of how humans learn. Our evolution as a species is tightly linked to the ability to exchange learned knowledge with one another. It follows that experience sharing (ES) between autonomous and independent agents could become the key to accelerate learning in cooperative multiagent settings. We investigate if randomly selecting experiences to share can increase the performance of deep reinforcement learning agents, and propose three new methods for selecting experiences to accelerate the learning process. Firstly, we introduce Focused ES, which prioritizes unexplored regions of the state space. Secondly, we present Prioritized ES, in which temporal-difference error is used as a measure of priority. Finally, we devise Focused Prioritized ES, which combines both previous approaches. The methods are empirically validated in a control problem. While sharing randomly selected experiences between two Deep Q-Network agents shows no improvement over a single agent baseline, we show that the proposed ES methods can successfully outperform the baseline. In particular, the Focused ES accelerates learning by a factor of 2, reducing by 51% the number of episodes required to complete the task.

The traffic assignment problem (TAP) plays a key role in the context of efficient urban mobility. The TAP can be approached from various perspectives. One of the fundamental models to solve the TAP is the so-called User Equilibrium (UE), which assumes that drivers behave rationally aiming at minimising their travel costs. However, this is a complex optimisation problem. To this regard, in this paper we propose the use of the GRASP metaheuristic to provide approximate solutions to the UE. The path relinking mechanism is also used to increase the coverage of the solutions space. We advance the state-of-the-art by proposing a novel modelling for the TAP, through which one can adjust the granularity of the search space, thus making a more efficient, directed local search. We also devise an efficient assignment evaluation scheme that avoids redundant computations during the local search process. Additionally, we develop a novel greedy procedure for generating enhanced initial solutions for the GRASP algorithm. Based on experiments, we demonstrate that our approach outperforms classical algorithms, providing solutions that are significantly closer to the UE. Moreover, our empirical results show that the stability and fairness levels achieved by our approach are considerably better than those achieved by other methods.

Solving the traffic assignment problem (TAP) is an important step towards an efficient usage of the traffic infrastructure. A fundamental assignment model is the so-called User Equilibrium (UE), which may turn into a complex optimisation problem. In this paper, we present the use of the GRASP metaheuristic to approximate the UE of the TAP. A path relinking mechanism is also employed to promote a higher coverage of the search space. Moreover, we propose a novel performance evaluation function, which measures the number of vehicles that have an incentive to deviate from the routes to which they were assigned. Through experiments, we show that our approach outperforms classical algorithms, providing solutions that are, on average, significantly closer to the UE. Furthermore, when compared to classical methods, the fairness achieved by our assignments is considerably better. These results indicate that our approach is efficient and robust, producing reasonably stable assignments.