The rapid advancement of large language models has significantly enhanced the quality of AI-generated text, making it increasingly difficult for detection systems to distinguish from human-written content. Existing detection methods, such as statistical, linguistic, machine learning, and deep learning approaches, often exhibit a decline in performance when applied to new or previously unseen large language models. Additionally, they tend to become outdated due to their static frameworks and inability to adapt to emerging patterns in generative text. To address this limitation, we introduce a novel dynamic fusion framework that integrates multi-scale feature fusion to capture diverse text patterns and employs continual learning with Elastic Weight Consolidation (EWC) to adapt to new models while mitigating catastrophic forgetting. This is the first attempt, to the best of our knowledge, to develop such a dynamic framework for AI-generated text detection. Evaluated on the TuringBench and DeepfakeTextDetect benchmark datasets, our framework achieves an average accuracy of 95.78% and 92.39%, outperforming the standard model by 5.88% and 7.98%, respectively, in distinguishing AI-generated from human-written text across various language generative model architectures. The continual learning ensures that the model remains adaptive and accurate over time, which is essential for practical applications in dynamic environments. This dynamic and adaptive approach paves the way for resilient AI-generated text detection systems capable of evolving alongside the rapidly advancing landscape of generative language technologies.

Efficient train scheduling on single-track railways represents a significant challenge due to operational constraints. Multiple trains share the same track and an optimal schedule must ensure that trains traveling in opposite directions do not collide while minimising delays. This paper proposes a novel approach to address this problem by formulating train scheduling as a distributed constraint satisfaction problem and applying a multi-agent system to solve it. We propose a simple, yet efficient, system in which agents cooperate to schedule trains on a single-track railway. The result shows that our system is reliable and fast in comparison to other popular approaches.

Abstract Deep learning models have been instrumental in extracting critical indicators for breast cancer diagnosis - the prevalent malignancy among women worldwide - from baseline magnetic resonance imaging. However, many existing models do not fully leverage the rich spatial information available in the 3D structure of medical imaging data, potentially overlooking important contextual details. This develops an explainable deep learning framework for classifying breast cancer that leverages the complete 3D and provides classification results alongside visual explanations of the decision-making process. The preprocessing pipeline is fed with 3D sequences containing ‘tumour’ and ‘non-tumour’ regions. It includes a 3D Adaptive Unsharp Mask (AUM) filter to reduce noise and augment image class, followed by normalisation and data augmentation. Classification is then achieved by training an augmented ResNet150 model. Three explainable artificial intelligence (XAI) techniques, including Shapley Additive Explanations, 3D Gradient-Weighted Class Activation Mapping, and Contextual Importance and Utility, are employed to provide improved interpretability. The model demonstrates state-of-the-art performance over the QIN-BREAST dataset, achieving testing accuracies of 98.861% for ‘tumours’ and 99.447% for ‘non-tumours’, as well as over the Duke Breast Cancer Dataset, where it achieves 99.104% for ‘tumours’ and 99.753% for ‘non-tumours’, while offering enhanced interpretability through XAI methods.

ABSTRACT This study presents a novel approach to crop mapping using remotely sensed satellite images. It addresses the significant classification modelling challenges, including (1) the requirements for extensive labelled data and (2) the complex optimisation problem for selection of appropriate temporal windows in the absence of prior knowledge of cultivation calendars. We compare the lightweight Dynamic Time Warping (DTW) classification method with the heavily supervised Convolutional Neural Network ‐ Long Short‐Term Memory (CNN‐LSTM) using high‐resolution multispectral optical satellite imagery (3 m/pixel). Our approach integrates effective practical preprocessing steps, including data augmentation and a data‐driven optimisation strategy for the temporal window, even in the presence of numerous crop classes. Our findings demonstrate that DTW, despite its lower data demands, can match the performance of CNN‐LSTM through our effective preprocessing steps while significantly improving runtime. These results demonstrate that both CNN‐LSTM and DTW can achieve deployment‐level accuracy and underscore the potential of DTW as a viable alternative to more resource‐intensive models. The results also prove the effectiveness of temporal windowing for improving runtime and accuracy of a crop classification study, even with no prior knowledge of planting timeframes.

We argue that results produced by a heuristic optimisation algorithm cannot be considered reproducible unless the algorithm fully specifies what should be done with solutions generated outside the domain, even in the case of simple bound constraints. Currently, in the field of heuristic optimisation, such specification is rarely mentioned or investigated due to the assumed triviality or insignificance of this question. Here, we demonstrate that, at least in algorithms based on Differential Evolution, this choice induces notably different behaviours in terms of performance, disruptiveness, and population diversity. This is shown theoretically (where possible) for standard Differential Evolution in the absence of selection pressure and experimentally for the standard and state-of-the-art Differential Evolution variants, on a special test function and the BBOB benchmarking suite, respectively. Moreover, we demonstrate that the importance of this choice quickly grows with problem dimensionality. Differential Evolution is not at all special in this regard-there is no reason to presume that other heuristic optimisers are not equally affected by the aforementioned algorithmic choice. Thus, we urge the heuristic optimisation community to formalise and adopt the idea of a new algorithmic component in heuristic optimisers, which we refer to as the strategy of dealing with infeasible solutions. This component needs to be consistently: (a) specified in algorithmic descriptions to guarantee reproducibility of results, (b) studied to better understand its impact on an algorithm's performance in a wider sense (i.e., convergence time, robustness, etc.), and (c) included in the (automatic) design of algorithms. All of these should be done even for problems with bound constraints.

Addressing climate change represents one of the most pressing challenges for organisations in developing nations. This is particularly relevant for companies navigating the shift towards a low‐carbon economy. This research leverages artificial intelligence (AI) methodologies to evaluate the financial implications of climate transition risks, encompassing both direct and indirect energy usage, including expenditures on electricity and fossil fuels. Advanced machine learning (ML) and deep learning (DL) models are employed to predict electricity and diesel consumption trends along with their associated costs. Findings from this study indicate an average prediction accuracy of 90.36%, underscoring the value of these tools in supporting organisational decision making related to climate transition risks. The study lays a foundation for comprehending not only the added costs linked to climate risks but also the potential advantages of transitioning to a low‐carbon economy, particularly from an energy‐focused perspective. Additionally, the proposed climate transition risk adjustment factor offers a framework for visualising the financial impacts of scenarios outlined by the Network for Greening the Financial System.

New contributions in the field of iterative optimisation heuristics are often made in an iterative manner. Novel algorithmic ideas are not proposed in isolation, but usually as extensions of a preexisting algorithm. Although these contributions are often compared to the base algorithm, it is challenging to make fair comparisons between larger sets of algorithm variants. This happens because even small changes in the experimental setup, parameter settings, or implementation details can cause results to become incomparable. Modular algorithms offer a way to overcome these challenges. By implementing the algorithmic modifications into a common framework, many algorithm variants can be compared, while ensuring that implementation details match in all versions.

Recently, new technologies have been developed that allow physical and virtual space to converge. We reviewed a range of innovative technologies that enable immersive and 3D interaction, which we believe are of particular interest to apply Universal Design for Learning (UDL) principles in teaching and learning practices, with a particular interest in higher education. We found limitations related to hardware and interactive systems that cannot be customised to meet the needs of all different students and some users may be marginalised. We draw attention to problems that lead to the risk of intentional exclusion while highlighting relevant inclusion opportunities. The main contribution of this paper is to present a selection of use cases to discuss how software applications can be designed to meet the guidelines for UDL, and improve the accessibility of 3D interaction for innovative Virtual Learning Environments (VLEs).

The financial impacts of the consequences of climate change on organisations are not always clear. Therefore, for many organisations, assessing the potential impact of climate risk remains a challenge. This piece of research is geared towards selecting the best artificial intelligence technique for modelling and forecasting the operational costs related to climate transition risk, such as direct or indirect utility consumption, including the costs of electricity and fossil fuels. These costs can be affected, especially by regulatory and market risks that are part of transition risk. We train 17 deep learning and machine learning models to forecast electricity and diesel consumption indicators, as well as the corresponding prices of these variables. Our results show an average prediction accuracy of 90.36% and emphasise the importance of using such decision support tools to analyse the financial impacts of climate risks within organisations. This calls for organisations to improve data collection and availability, as performing this kind of analysis is promising but currently remains a challenge due to the scarcity of available information.

In the last three decades, the field of computational intelligence has seen a profusion of population‐based metaheuristics applied to a variety of problems, where they achieved state‐of‐the‐art results. This remarkable growth has been fuelled and, to some extent, exacerbated by various sources of inspiration and working philosophies, which have been thoroughly reviewed in several recent survey papers. However, the present survey addresses an important gap in the literature. Here, we reflect on a systematic categorisation of what we call “lightweight” metaheuristics, i.e., optimisation algorithms characterised by purposely limited memory and computational requirements. We focus mainly on two classes of lightweight algorithms: single‐solution metaheuristics and “compact” optimisation algorithms. Our analysis is mostly focused on single‐objective continuous optimisation. We provide an updated and unified view of the most important achievements in the field of lightweight metaheuristics, background concepts, and most important applications. We then discuss the implications of these algorithms and the main open questions and suggest future research directions.

Rain can significantly impair the driver’s sight and affect his performance when driving in wet conditions. Evaluation of driver visibility in harsh weather, such as rain, has garnered considerable research since the advent of autonomous vehicles and the emergence of intelligent transportation systems. In recent years, advances in computer vision and machine learning led to a significant number of new approaches to address this challenge. However, the literature is fragmented and should be reorganised and analysed to progress in this field. There is still no comprehensive survey article that summarises driver visibility methodologies, including classic and recent data‐driven/model‐driven approaches on the windshield in rainy conditions, and compares their generalisation performance fairly. Most ADAS and AD systems are based on object detection. Thus, rain visibility plays a key role in the efficiency of ADAS/AD functions used in semi‐ or fully autonomous driving. This study fills this gap by reviewing current state‐of‐the‐art solutions in rain visibility estimation used to reconstruct the driver’s view for object detection‐based autonomous driving. These solutions are classified as rain visibility estimation systems that work on (1) the perception components of the ADAS/AD function, (2) the control and other hardware components of the ADAS/AD function, and (3) the visualisation and other software components of the ADAS/AD function. Limitations and unsolved challenges are also highlighted for further research.

Computational Intelligence (CI), originally represented by the three subjects of Evolutionary Computation (EC), Fuzzy Logic (FL) and Neural Networks (NNs), has significantly evolved to date and is ever more embedded in both software platforms and hardware devices forming intelligent systems capable of self-adaptation, decision-making and problem-solving.With a quick inspection of the scientific literature in Computer Science, one can indeed notice a significant expansion in the range of available CI tools, with, for example, modern EC optimisers making use of surrogate models (which can be based on NNs), or being used to evolve both topology and hyperparameters of neural systems.The latter systems have also grown significantly and currently offer numerous kinds of networks from, for example, recurrent, through convolutional to Generative/Adversarial deep NNs.These highly interconnected and high-level algorithms are becoming ubiquitous as their applicability has widened and grown to traverse many disciplines and application domains.In the past, the technological fields that benefited the most from applying CI techniques were in engineering, such as system control and design, robotics, telecommunication and so forth.However, the application scope of modern CI methods has widened significantly, thus making it possible to analyse large data sets, manipulate images and videos, extract sentiment and relevant information from plain text and audio recordings.Hence, modern CI turns out to be helpful in many areas which strongly impact our society, for example, medicine, finance, education, intelligent transportation, sustainability and so forth, where it is key to analyse available data, optimise processes and provide systems with extra capabilities.If placed in the right context, CI has then the potential of generating societal impact beyond enabling technological advancement per se.It can now support the deployment of technology to optimise not only the financial viability but as well the usability and benefit to the public.State-of-the-art optimisation has become focused on sustainability and waste rather than profit or cost reduction; now optimisation is critical to address the compromise between protecting society and the economic activities of small stockholders, and not just the large scale businesses.In this light, this special issue has gathered recent advances in CI addressing relevant research questions leading to societal impact and calling for the design of more intelligent systems enhancing our society in the future.

The unfolding coronavirus (COVID-19) pandemic has highlighted the global need for robust predictive and containment tools and strategies. COVID-19 continues to cause widespread economic and social turmoil, and while the current focus is on both minimising the spread of the disease and deploying a range of vaccines to save lives, attention will soon turn to future proofing. In line with this, this paper proposes a prediction and containment model that could be used for pandemics and natural disasters. It combines selective lockdowns and protective cordons to rapidly contain the hazard while allowing minimally impacted local communities to conduct "business as usual" and/or offer support to highly impacted areas. A flexible, easy to use data analytics model, based on Self Organising Maps, is developed to facilitate easy decision making by governments and organisations. Comparative tests using publicly available data for Great Britain (GB) show that through the use of the proposed prediction and containment strategy, it is possible to reduce the peak infection rate, while keeping several regions (up to 25% of GB parliamentary constituencies) economically active within protective cordons.

Benchmarking heuristic algorithms is vital to understand under which conditions and on what kind of problems certain algorithms perform well. Most benchmarks are performance based, to test algorithm performance under a wide set of conditions. There is also resource- and behavior-based benchmarks to test the resource consumption and the behavior of algorithms. In this article, we propose a novel behavior-based benchmark toolbox: BIAS (Bias in algorithms, structural). This toolbox can detect structural bias (SB) per dimension and across dimension-based on 39 statistical tests. Moreover, it predicts the type of SB using a random forest model. BIAS can be used to better understand and improve existing algorithms (removing bias) as well as to test novel algorithms for SB in an early phase of development. Experiments with a large set of generated SB scenarios show that BIAS was successful in identifying bias. In addition, we also provide the results of BIAS on 432 existing state-of-the-art optimization algorithms showing that different kinds of SB are present in these algorithms, mostly toward the center of the objective space or showing discretization behavior. The proposed toolbox is made available open-source and recommendations are provided for the sample size and hyper-parameters to be used when applying the toolbox on other algorithms.

Developments in artificial intelligence can be leveraged to support the diagnosis of degenerative disorders, such as epilepsy and Parkinson's disease. This study aims to provide a software solution, focused initially towards Parkinson's disease, which can positively impact medical practice surrounding degenerative diagnoses. Through the use of a dataset containing numerical data representing acoustic features extracted from an audio recording of an individual, it is determined if a neural approach can provide an improvement over previous results in the area. This is achieved through the implementation of a feedforward neural network and a layer recurrent neural network. By comparison with the state-of-the-art, a Bayesian approach providing a classification accuracy benchmark of 87.1%, it is found that the implemented neural networks are capable of average accuracy of 96%, highlighting improved accuracy for the classification process. The solution is capable of supporting the diagnosis of Parkinson's disease in an advisory capacity and is envisioned to inform the process of referral through general practice.

This article presents an intelligent system using deep learning algorithms and the transfer learning approach to detect oil palm units in multispectral photographs taken with unmanned aerial vehicles. Two main contributions come from this piece of research. First, a dataset for oil palm units detection is carefully produced and made available online. Although being tailored to the palm detection problem, the latter has general validity and can be used for any classification application. Second, we designed and evaluated a state-of-the-art detection system, which uses a convolutional neural network to extract meaningful features, and a classifier trained with the images from the proposed dataset. Results show outstanding effectiveness with an accuracy peak of 99.5% and a precision of 99.8%. Using different images for validation taken from different altitudes the model reached an accuracy of 97.5% and a precision of 98.3%. Hence, the proposed approach is highly applicable in the field of precision agriculture.

Lethal Wilt (Marchitez Letal) is a disease which affects Etaeis Guineensis, a plant used in the production of palm oil. The disease is increasingly common but the spatial-dynamics of the infection spread remain poorly understood. It is particularly dangerous due to the speed at which it spreads and the speed at which infected plants show symptoms and die. Early identification, or even better, accurate prediction of areas at high risk of infection can slow the spread of the disease and limit crop waste. This study is based on data collected over a five-year period from an affected plantation in Colombia. The aim of the study is to analyse the collected data to better understand how the disease spreads and then to model the behaviour. Based on insights from the initial analysis a multi-agent-based system is proposed to model the pattern of infection. The model is comprised of two steps; first Kernel Density Estimation is used to create an estimation of the distribution from which newly infected plants are drawn and this density estimation is then used to direct agents on a biased-walk of the surrounding areas. Results show that the model can approximate the behaviour of the disease and can predict areas which are at high risk of future infection.

Real-world problems such as landmine detection require multiple sources of information to reduce the uncertainty of decision-making. A novel approach to solve these problems includes distributed systems, as presented in this work based on hardware and software multi-agent systems. To achieve a high rate of landmine detection, we evaluate the performance of a trained system over the distribution of samples between training and validation sets. Additionally, a general explanation of the data set is provided, presenting the samples gathered by a cooperative multi-agent system developed for detecting improvised explosive devices. The results show that input samples affect the performance of the output decisions, and a decision-making system can be less sensitive to sensor noise with intelligent systems obtained from a diverse and suitably organised training set.

Robust optimization Over Time (ROOT) is a new method of solving Dynamic optimization Problems in respect to choosing a robust solution, that would last over a number of environment changes, rather than the approach that chooses the optimal solution at every change. ROOT methods currently show that ROOT can be solved by predicting an individual fitness for a number of future environment changes. In this work, a benchmark problem based on the Modified Moving Peaks Benchmark (MMPB) is proposed that includes an attractor heuristic, that guides optima to a determined location in the environment, resulting in a more predictable optimum. We study a number of time series forecasting methods to test different prediction methods of future fitness values in a ROOT method. Four time series regression techniques are considered as the prediction method: Linear and Quadratic Regression, an Autoregressive model, and Support Vector Regression. We find that there is not much difference in choosing a simple Linear Regression to more advanced prediction methods. We also suggest that current benchmark problems that cannot be predicted will deceive the optimizer and ROOT framework as the peaks may move using a random walk. Results show an improvement in comparison with MMPB used in most ROOT studies.

This paper investigates whether optimisation methods with the population made up of one solution can suffer from structural bias just like their multisolution variants. Following recent results highlighting the importance of choice of strategy for handling solutions generated outside the domain, a selection of single solution methods are considered in conjunction with several such strategies. Obtained results are tested for the presence of structural bias by means of a traditional approach from literature and a newly proposed here statistical approach. These two tests are demonstrated to be not fully consistent. All tested methods are found to be structurally biased with at least one of the tested strategies. Confirming results for multisolution methods, it is such strategy that is shown to control the emergence of structural bias in single solution methods. Some of the tested methods exhibit a kind of structural bias that has not been observed before.

This study combines two novel deterministic methods with a Convolutional Neural Network to develop a machine learning method that is aware of directionality of light in images. The first method detects shadows in terrestrial images by using a sliding-window algorithm that extracts specific hue and value features in an image. The second method interpolates light-sources by utilising a line-algorithm, which detects the direction of light sources in the image. Both of these methods are single-image solutions and employ deterministic methods to calculate the values from the image alone, without the need for illumination-models. They extract real-time geometry from the light source in an image, rather than mapping an illuminationmodel onto the image, which are the only models used today. Finally, those outputs are used to train a Convolutional Neural Network. This displays greater accuracy than previous methods for shadow detection and can predict light source-direction and thus orientation accurately, which is a considerable innovation for an unsupervised CNN. It is significantly faster than the deterministic methods. We also present a reference dataset for the problem of shadow and light direction detection.

The ability for grocery retailers to have a single view of customers across all their grocery purchases remains elusive and has become increasingly important in recent years (especially in the United Kingdom) where competition has intensified, shopping habits and demographics have changed and price sensitivity has increased following the 2008 recession. Numerous studies have been conducted on understanding independent items that are frequently bought together (association rule mining/frequent itemsets) with several measures proposed to aggregate item support and rule confidence with varying levels of accuracy as these measures are highly context dependent. Uninorms were used as an alternative measure to aggregate support and confidence in analysing market basket data using the UK grocery retail sector as a case study. Experiments were conducted on consumer panel data with the aim of comparing the uninorm against three other popular measures (Jaccard, Cosine and Conviction). It was found that the uninorm outperformed other models on its adherence to the fundamental monotonicity property of support in market basket analysis (MBA). Future work will include the extension of this analysis to provide a generalised model for market basket analysis.

We propose Multi-Strategy Coevolving Aging Particles (MS-CAP), a novel population-based algorithm for black-box optimization. In a memetic fashion, MS-CAP combines two components with complementary algorithm logics. In the first stage, each particle is perturbed independently along each dimension with a progressively shrinking (decaying) radius, and attracted towards the current best solution with an increasing force. In the second phase, the particles are mutated and recombined according to a multi-strategy approach in the fashion of the ensemble of mutation strategies in Differential Evolution. The proposed algorithm is tested, at different dimensionalities, on two complete black-box optimization benchmarks proposed at the Congress on Evolutionary Computation 2010 and 2013. To demonstrate the applicability of the approach, we also test MS-CAP to train a Feedforward Neural Network modeling the kinematics of an 8-link robot manipulator. The numerical results show that MS-CAP, for the setting considered in this study, tends to outperform the state-of-the-art optimization algorithms on a large set of problems, thus resulting in a robust and versatile optimizer.

The super-fit scheme, consisting of injecting an individual with high fitness into the initial population of an algorithm, has shown to be a simple and effective way to enhance the algorithmic performance of the population-based algorithm. Whether the super-fit individual is based on some prior knowledge on the optimization problem or is derived from an initial step of pre-processing, e.g. a local search, this mechanism has been applied successfully in various examples of evolutionary and swarm intelligence algorithms. This paper presents an unconventional application of this super-fit scheme, where the super-fit individual is obtained by means of the Covariance Adaptation Matrix Evolution Strategy (CMA-ES), and fed to a single solution local search which perturbs iteratively each variable. Thus, compared to other super-fit schemes, the roles of super-fit individual generator and global optimizer are switched. To prevent premature convergence, the local search employs a re-sampling mechanism which inherits parts of the best individual while randomly sampling the remaining variables. We refer to such local search as Re-sampled Inheritance Search (RIS). Tested on the CEC 2013 optimization benchmark, the proposed algorithm, named CMA-ES-RIS, displays a respectable performance and a good balance between exploration and exploitation, resulting into a versatile and robust optimization tool.

This paper proposes an algorithm to solve the CEC2013 benchmark. The algorithm, namely Super-fit Multicriteria Adaptive Differential Evolution (SMADE), is a Memetic Computing approach based on the hybridization of two algorithmic schemes according to a super-fit memetic logic. More specifically, the Covariance Matrix Adaptive Evolution Strategy (CMAES), run at the beginning of the optimization process, is used to generate a solution with a high quality. This solution is then injected into the population of a modified Differential Evolution, namely Multicriteria Adaptive Differential Evolution (MADE). The improved solution is super-fit as it supposedly exhibits a performance a way higher than the other population individuals. The super-fit individual then leads the search of the MADE scheme towards the optimum. Unimodal or mildly multimodal problems, even when non-separable and ill-conditioned, tend to be solved during the early stages of the optimization by the CMAES. Highly multi-modal optimization problems are efficiently tackled by SMADE since the MADE algorithm (as well as other Differential Evolution schemes) appears to work very well when the search is led by a super-fit individual.

This paper introduces two lightweight variants of ISPO, a Single Particle Optimization algorithm recently proposed in the literature. The goal of this work is to improve upon the performance of the original ISPO, still bearing in mind its admirable algorithmic simplicity. The first variant, namely ISPOrestart, combines in a memetic fashion the logics of ISPO with a partial restart mechanism similar to the binomial crossover typically used in Differential Evolution. The second variant, named VISPO, builds on top of the restart process a very simple learning stage which tries to adapt the algorithm behaviour to the (non)-separability of the problem. Numerical results obtained on three complete optimization benchmarks show that not only the two algorithms are able to improve, incrementally, upon the performance of ISPO, but also they show respectable performance in comparison with modern complex state-of-the-art methods, especially when the problem dimensionality increases.