Tensor networks provide efficient representations for high dimensional data, although selecting suitable neural architectures for their integration remains challenging. Neural Architecture Search (NAS) offers a promising solution by automating architecture design through structured search spaces and algorithmic optimization, often matching or surpassing human-crafted models. In this work, we evaluate the selected NAS methods on CIFAR-10 using common metrics including clean and adversarial accuracy, F1-score, parameter count, and search cost. The experimental results show that DARTS delivers the highest clean accuracy with minimal computational load, whereas HA-ENAS achieves the strongest adversarial robustness due to its multi-objective evolutionary strategy. REP consistently enhances robustness within differentiable NAS frameworks, especially on structured benchmarks, while NAS-RL demonstrates significant scalability limitations and limited architectural diversity. Overall, findings confirm that no single NAS method dominates across all evaluation criteria. Instead, effectiveness depends on the trade-off between robustness, efficiency, and architectural complexity. The study concludes that REP provides the most balanced performance, DARTS excels in efficiency, and HA-ENAS is best suited for robustness-oriented applications, offering a practical foundation for future NAS research in tensor-network–based model design.

This work detects predators in extensive breeding farms, especially those dedicated to free-range chicken production. In this environment, the main threats are birds of prey and poachers. The proposal focuses on the first case to study how using deep learning models can help detect and label images. Specifically, we evaluate the performance of several object detection models on a dataset focused on birds of prey in outdoor farm settings. The study compares architectures from two families: Faster R-CNN and YOLO. Results show notable differences in performance between model families. Faster R-CNN shows a better performance than YOLO. This considerable difference marks the importance of selecting the model to be used, especially in cases such as this one, where the objects to be detected and classified are small in size.

This work presents a novel tensor network algorithm for solving Quadratic Unconstrained Binary Optimization (QUBO) problems, Quadratic Unconstrained Discrete Optimization (QUDO) problems, and Tensor Quadratic Unconstrained Discrete Optimization (T-QUDO) problems. The proposed algorithm is based on the MeLoCoToN methodology, which solves combinatorial optimization problems by employing superposition, imaginary time evolution, and projective measurements. Additionally, two different approaches are presented to solve QUBO and QUDO problems with k-neighbors interactions in a lineal chain, one based on 4-order tensor contraction and the other based on matrix-vector multiplication, including sparse computation and a new technique called "Waterfall". Furthermore, the performance of both implementations is compared with a quadratic optimization solver to demonstrate the performance of the method, showing advantages in several problem instances.

This study focuses on an industrial facility’s time-slot allotment (TSA) problem, where loading and unloading docks are assigned to the incoming lorries, reducing the number of them waiting for service. Several constraints apply to the different dock stations, including disparate timetables and task duration, various capacities of simultaneous services, etc. Evolutionary algorithms cope with the enormous variability of the combinatorial problem, maximizing the number of accepted lorries while decreasing the queue at the entrance. Interestingly, the problem’s structure led to unconventional operator probabilities, which also analyzes the evolutionary techniques and operators included in this study. Comparative analysis with state-of-the-art methods highlights the algorithm’s effectiveness, though computational demands rise with population size.

This paper presents a controlled stochastic pipeline for generating realistic acoustic scenes for human-related event classification under real background interference. As a first contribution, the pipeline is introduced as a general tool for constructing reproducible benchmarks with configurable event presence, class balance, and signal degradation conditions. This enables systematic analysis of background dominance, imbalance, and signal-to-noise effects. As a second contribution, the paper reports an experimental deployment of the pipeline using 18 human-event classes from FSD50K and long ambient recordings collected at five real industrial locations of TSK. In this case study, the generator is instantiated under three class-imbalance rates—realist, semibalanced, and balanced—producing 1-second mixtures with SNRs uniformly sampled from -30 to +30 dB. The generated benchmarks are used to compare nine shallow classifiers trained on 88 eGeMAPSv02 functional descriptors and two deep architectures trained on log-mel spectrograms, all evaluated under a shared nested cross-validation protocol. Results show that the generation regime strongly conditions performance. Deep models outperform shallow baselines, with AudioCNN trained with categorical cross-entropy obtaining the best overall results. Performance improves under less sparse regimes and higher SNR values, although confusions remain among acoustically similar speech-related classes. These findings support the proposed pipeline as both a benchmark-generation tool and a practical framework for robust acoustic event classification analysis.

In this paper, we present two tensor network quantum-inspired algorithms to solve the knapsack and the shortest path problems, and enables to solve some of its variations. These methods provide an exact equation which returns the optimal solution of the problems. As in other tensor network algorithms for combinatorial optimization problems, the method is based on imaginary time evolution and the implementation of restrictions in the tensor network. In addition, we introduce the use of symmetries and the reutilization of intermediate calculations, reducing the computational complexity for both problems. To show the efficiency of our implementations, we carry out some performance experiments and compare the results with those obtained by other classical algorithms.

Anomaly detection is a key task across domains such as industry, healthcare, and cybersecurity. Many real-world anomaly detection problems involve analyzing multiple features over time, making time series analysis a natural approach for such problems. While deep learning models have achieved strong performance in this field, their trend to exhibit high energy consumption limits their deployment in resource-constrained environments such as IoT devices, edge computing platforms, and wearables. To address this challenge, this paper introduces the \textit{Vacuum Spiker algorithm}, a novel Spiking Neural Network-based method for anomaly detection in time series. It incorporates a new detection criterion that relies on global changes in neural activity rather than reconstruction or prediction error. It is trained using Spike Time-Dependent Plasticity in a novel way, intended to induce changes in neural activity when anomalies occur. A new efficient encoding scheme is also proposed, which discretizes the input space into non-overlapping intervals, assigning each to a single neuron. This strategy encodes information with a single spike per time step, improving energy efficiency compared to conventional encoding methods. Experimental results on publicly available datasets show that the proposed algorithm achieves competitive performance while significantly reducing energy consumption, compared to a wide set of deep learning and machine learning baselines. Furthermore, its practical utility is validated in a real-world case study, where the model successfully identifies power curtailment events in a solar inverter. These results highlight its potential for sustainable and efficient anomaly detection.

Digitalization of electricity infrastructures has accelerated the evolution of Smart Grids, with Energy Communities representing a pivotal advancement in decentralized energy management. This study introduces a decision-making framework that optimizes energy exchanges through a dynamic evaluation process, adapting in real time to participants' preferences on energy usage, considered as User Policies (UPs). To achieve optimal energy coordination, the proposed methodology incorporates demand response strategies, enabling participants to regulate consumption through controllable loads and storage assets. A structured multi-objective optimization approach formalizes decision-making, embedding the five key User Policies (i.e. economic efficiency, peak power mitigation, sustainability, comfort, and energy efficiency) into an evaluation function that quantifies prosumer preferences. Fairness and arbitrage mechanisms are employed to maintain system equilibrium and prevent inefficiencies in energy trading to achieve the goal of long-term sustainability. The framework dynamically reacts to fluctuations, optimizing energy distribution while maintaining system stability and long-term adaptability. Computationally, the framework leverages genetic optimization techniques to enhance prediction accuracy and decision-making efficiency. The results demonstrate significant improvements in energy allocation, peak demand reduction, and financial optimization, validating the framework's effectiveness. By systematically aligning individual and collective energy objectives, this methodology provides a scalable and adaptive approach to decentralized energy management, reinforcing the resilience and intelligence of next-generation Smart Grids.

Function optimization represents a remarkable challenge in industry and society, aiming to find reasonable solutions-even if they are suboptimal-for everyday problems. Metaheuristics drive the optimization search towards the goals using a specific algorithm inspired by different concepts: from industrial processes to the behaviour of living beings in nature, from mathematical ideas to physics notions. This research proposes a new metaheuristic inspired by the Slime Mould and its foraging behaviours. On the one hand, an exploitation stage mimics the greedy amoeba's conduct when food is plenty. On the other hand, an exploration stage copies the fruity aggregation of the cells and the subsequent spore dissemination. This study compares the most cited metaheuristics and the Slime Mould Optimization in two different experimentation stages: on the one hand, the optimization of standard benchmarking functions; on the other hand, solving the robot path planning problem. Moreover, a hybridization of the SMO and the WOA is presented, which keeps the SMO's convergence speed and the WOA's good performance in finding the best solutions.

Stress significantly impacts individuals, particularly in professions like nursing and driving, leading to severe health risks and accidents. Accurate stress measurement is critical for effective interventions, yet research is hindered by incomplete datasets and inconsistent methodologies, slowing the development of reliable predictive models. This paper introduces a framework for cross-contextual stress prediction, enabling the generation of general stress prediction models adaptable to specific domain challenges. The methodology leverages two general daily life datasets and three domain-specific datasets, employing steps such as dataset selection, feature extraction, significant feature identification, feature preprocessing, fine-tuning, domain adaptation, and application to specific contexts. Through this framework, key vital signs were identified as significant predictors of stress, including electrocardiography (ECG), heart rate (HR), heart rate variability (HRV) - low frequency (LF), electrodermal activity (EDA), body temperature (TEMP), and skin conductance response (SCR). The experiments conducted include: 1) Utilizing HR and HRV-LF through domain adaptation from general to automobile driving datasets; 2) Applying EDA, HR, and TEMP from general to specific nurse activity datasets; and 3) Adapting ECG, HR, and TEMP from general to automobile driving datasets. Results demonstrate the potential of the proposed framework for cross-contextual stress prediction, with HR and HRV-LF identified as pivotal features. When applied to target datasets specific to stress scenarios, the model achieved a 62% F1 score, demonstrating the effectiveness of the feature-based Correlation Alignment (CORAL) technique combined with Random Forest models in transferring learned knowledge across domains. These findings highlight the robustness of the approach in adapting general stress prediction models to specific contexts, paving the way for real-world applications such as stress monitoring in driving and nursing during high-stress periods like COVID-19.

In this work, we deal with the Multirobot Path Planning (MPP) problem, one of the most extended problems in the planning literature due to its wide application to the real world. The counterpart of this work is the proposal of a new framework for the MPP problem that better represents its application to a natural-world environment, as opposed to other approaches with a more theoretical and academic demonstration. The new approach is inspired by a scenario in which a set of robots must move through a closed environment to complete a series of tasks. Along with the proposal of the new statement, a solution method is proposed with an anytime algorithm in which the route can be requested for a robot present in the scenario and planned for it, considering the presence of the rest of the robots. This algorithm is designed from its real application, so it is configurable according to the needs and prioritisation of the environment in which it moves. The results show that the proposed method is very efficient with the established problem of planning all the requested routes without collisions and fast response times .

This article reviews the metaheuristics published in the literature, emphasizing their usefulness in solving complex optimization problems. The review highlights inspiration's relevance in the metaheuristics design, being the main classification in multiple taxonomies existing in the literature. After reviewing the state of the art, six of the most relevant metaheuristics used to solve problems of various types (engineering, logistics, economics, data science, ...) were selected. This selection of metaheuristics will be subjected to an analysis of their performance using a set of problems selected from different authors. The problems selected for this analysis include problems with a single minimum or multiple minima, different sizes in terms of dimensions, and different types of mathematical functions such as polynomial, trigonometric, or exponential. The analysis offers a discussion of which scenarios are the best for each metaheuristic, analyzing aspects such as the ability of metaheuristics to explore and escape local minima. The article concludes by summarizing which metaheuristic is best for each type of problem. Keywords: Metaheuristics, benchmark, optimization problems, biological-based metaheuristics

The deployment and inference of Deep Learning models are investigated in this article on seven different Edge Computing (EC) devices. This study addresses the features, performance and limitations of the NVIDIA Jetson Orin NX and Nano, Google Coral DevBoard and USB, Intel Neural Compute Stick 2, NXP i.MX8 Plus and Xilinx Zynq UltraScale+ MPSoC ZCU104 on Deep Learning inference. Fully Connected, Convolutional and Long Short-Term Memory (LSTM) neural networks are implemented to test these EC devices. The benchmarking focuses on the performance metrics: inference latency, increase in error metric, and power consumption. The results show considerable variability among devices, with the ZCU104 and Jetson Orin achieving the lowest latencies across most models without any increase in the error metric. At the same time, Coral devices exhibit increased latency and error for complex convolutional models. NVIDIA Jetson devices, ZCU104 and Neural Compute Stick 2 are the only devices that support LSTM inference. The study also highlights differences in power consumption, with USB accelerators being the most energy-efficient.

A computer-vision-based industrial algorithm is proposed in this study for the detection of the dimensions and the spatial positioning of fruit and vegetables on a conveyor belt for their movement to a packing machine with a robotic arm. The principal purpose of the algorithm is to identify the dimensions (length, width), and position (angle of inclination, and Cartesian coordinates) of the vegetable mesh netting without taking the product label into account. The proposed algorithm has four functions. A convolutive neuronal network model is applied for object detection, with which all objects are identified in the image while the product label is suppressed, so that recognition of the vegetable mesh netting and its dimensions is not impaired. Moreover, the Canny edge detector and the Border following algorithms are applied to perform image pre-processing and edge improvements, respectively, yielding optimal results with more clearly defined noise-free objects. The feature extraction function for dimensions yielded detection precision results of 99.60

In the current context of energy crisis, the energy sector is undergoing significant transformations towards a sustainable, competitive, and affordable energy landscape. Central to this transformation are Energy Communities (ECs), which have emerged as an ideal vehicle to facilitate the energy transition for small consumers. This paper evaluates the performance of three alternative methods (i.e. bill-sharing, price-based and surplus-based) to allocate costs and benefits within ECs. Specifically, we compare the distribution of savings generated by energy internal trading under each of the three allocation methods. The three allocation methods guarantee extracting the maximum economic surplus (i.e. savings) from the internal market, but one of them (bill-sharing) does not guarantee that participating in the internal trading is beneficial for every member. This is a major obstacle for its implementation since, frequently, some agents are worse off due to its participation in the community. The other two methods guarantee that participation is beneficial, but they differ in how savings are distributed. The distribution under price-based methods is influenced by the prices at which different members of the EC can buy and sell energy from the grid, while surplus-based methods distribute savings according to a criterion agreed by the EC members. Here we assume that they do it proportionally to the energy traded in the internal market. Keywords: energy communities, local energy market, allocation method

Metaheuristics are one of the most widespread solutions in the literature to address optimization problems. This application of metaheuristics makes them suitable for many problems in different fields of knowledge. However, the constant increase of new metaheuristics proposals has a problem; the comparison between different metaheuristics to determine which one to use is a very complex task for which it does not help that each author uses his way of evaluating his metaheuristics. In this work, we analyze and study different techniques used by authors of metaheuristics and evaluate their problems to propose a testing methodology that is objective, neutral, and fair for all metaheuristics evaluated. This methodology will also consider how to set the parameter optimization so that it is the same for all metaheuristics and problem instances to get as close as possible to an actual application of the metaheuristics. This new methodology will be tested in the experimentation by evaluating five metaheuristics widely known and extended in the literature.

The computational analysis of environmental sounds has garnered considerable attention in recent years, with a wide range of applications spanning media retrieval, hearing assistance, and biomonitoring systems. The two main tasks in this topic are acoustic scene classification and sound event detection. Current work presents the preliminary results of our first study of human spontaneous event identifications in real environments. The research will tackle the problem as an anomaly detection challenge, exploring the performance of semi-supervised and unsupervised shallow machine learning techniques on our sound datasets. Two sets of anomaly detection datasets have been created using our method of dataset fusion, combining a variety of acoustic scenes collected from a free sound library, as well as from different locations of our own real industrial company, and spontaneous human acoustic events from another free sound library. In addition, each dataset has been created following 7 Signal-Noise Ratio (SNR) challenges from -30dB up to +30dB, indicating that the semi-supervised algorithm Xtreme Boosting Based Outlier Detection, with 50

The detection of bacterial and viral microbes is pivotal for both human and animal well-being in public health services and veterinary care, but it traditionally requires time-consuming procedures and expert technicians. However, the rise of Machine Learning and Deep Learning has led to a surge in the application of new techniques that can perform bacterial and viral detection faster and at a lower cost. Yet, despite that success, Deep Learning approaches tend to have high energy demands, which can in some contexts limit their application, increasing both costs and environmental concerns. In this study, a new hybrid methodology, in which an Artificial Neural Network was combined with a more energy efficient Spiking Neural Network, was employed to develop a model able to classify 18 species of Eimeria parasites, affecting both rabbits and chickens, from microscope images. We show how significant energy savings can be obtained from SNN layers, while their use in the model can improve its performance.