Abstract Large language models (LLMs) have been prominent in various tasks, including text generation and summarisation. The applicability of LLMs to the generation of product reviews is gaining momentum, paving the way for the generation of movie reviews. In this study, we propose a framework that generates movie reviews using three LLMs (GPT-4o, DeepSeek-V3, and Gemini-2.0), and evaluate their performance by comparing the generated outputs with IMDb user reviews. We use movie subtitles and screenplays as input to the LLMs and investigate how they affect the quality of reviews generated. We review the LLM-based movie reviews in terms of vocabulary, sentiment polarity, similarity, and thematic consistency in comparison to IMDB user reviews. The results demonstrate that LLMs are capable of generating syntactically fluent and structurally complete movie reviews. Nevertheless, we find a noticeable gap in emotional richness and stylistic coherence between LLM-generated and IMDb reviews, suggesting that further refinement is needed to improve the overall quality of movie review generation. We finally conduct a survey-based human analysis, where participants distinguish between LLM and IMDb user reviews. The results show that LLM-generated reviews were generally difficult to distinguish from IMDB user reviews. We find that DeepSeek-V3 produced the most balanced reviews, closely matching IMDb reviews. GPT-4o overemphasised positive emotions, while Gemini-2.0 captured negative emotions better but showed excessive emotional intensity.

Recommender systems leveraging deep learning have significantly improved personalised item suggestions, yet single-criteria models often overlook the nuanced nature of user preferences. Multi-Criteria Recommender Systems (MCRS) address this by modeling multiple aspects (e.g., taste, appearance, location). However, existing deep learning approaches—especially those using shared embeddings or matrix factorization—struggle to capture complex structural and cross-criteria dependencies. To overcome these challenges, we propose a novel framework, D-MGAC (Dual Multiview Graph Attention and Contrastive learning), which formulates MCRS as a multi-edge bipartite graph and applies multiview dual graph attention mechanisms to model both local (per-criterion) and global (cross-criteria) interactions. Furthermore, we define anchor-based contrastive learning in both local and global views to refine the representation quality. Experiments on Yahoo!Movies and BeerAdvocate datasets demonstrate D-MGAC’s superiority, outperforming recent state-of-the-art models.

Pedestrian trajectory prediction plays an important role in autonomous driving systems and robotics. Recent work utilizing prominent deep learning models for pedestrian motion prediction makes limited a priori assumptions about human movements, resulting in a lack of explainability and explicit constraints enforced on predicted trajectories. We present a dynamics-based deep learning framework with a novel asymptotically stable dynamical system integrated into a Transformer-based model. We use an asymptotically stable dynamical system to model human goal-targeted motion by enforcing the human walking trajectory, which converges to a predicted goal position, and to provide the Transformer model with prior knowledge and explainability. Our framework features the Transformer model that works with a goal estimator and dynamical system to learn features from pedestrian motion history. The results show that our framework outperforms prominent models using five benchmark human motion datasets.

Drug repurposing (or repositioning) is the process of finding new therapeutic uses for drugs already approved by drug regulatory authorities (e.g., the Food and Drug Administration (FDA) and Therapeutic Goods Administration (TGA)) for other diseases. This involves analysing the interactions between different biological entities, such as drug targets (genes/proteins and biological pathways) and drug properties, to discover novel drug–target or drug–disease relations. Machine learning and deep learning models have successfully analysed complex heterogeneous data with applications in the biomedical domain, and have also been used for drug repurposing. This study presents a novel unsupervised machine learning framework that utilizes a graph-based autoencoder for multi-feature type clustering on heterogeneous drug data. The dataset consists of 438 drugs, of which 224 are under clinical trials for COVID-19 (category A). The rest are systematically filtered to ensure the safety and efficacy of the treatment (category B). The framework solely relies on reported drug data, including its pharmacological properties, chemical/physical properties, interaction with the host, and efficacy in different publicly available COVID-19 assays. Our machine-learning framework revealed three clusters of interest and provided recommendations featuring the top 15 drugs for COVID-19 drug repurposing, which were shortlisted based on the predicted clusters that were dominated by category A drugs. Our framework can be extended to support other datasets and drug repurposing studies with the availability of our open-source code.

Class imbalance (CI) in classification problems arises when the number of observations belonging to one class is lower than the other. Ensemble learning combines multiple models to obtain a robust model and has been prominently used with data augmentation methods to address class imbalance problems. In the last decade, a number of strategies have been added to enhance ensemble learning and data augmentation methods, along with new methods such as generative adversarial networks (GANs). A combination of these has been applied in many studies, and the evaluation of different combinations would enable a better understanding and guidance for different application domains. In this paper, we present a computational study to evaluate data augmentation and ensemble learning methods used to address prominent benchmark CI problems. We present a general framework that evaluates 9 data augmentation and 9 ensemble learning methods for CI problems. Our objective is to identify the most effective combination for improving classification performance on imbalanced datasets. The results indicate that combinations of data augmentation methods with ensemble learning can significantly improve classification performance on imbalanced datasets. We find that traditional data augmentation methods such as the synthetic minority oversampling technique (SMOTE) and random oversampling (ROS) are not only better in performance for selected CI problems, but also computationally less expensive than GANs. Our study is vital for the development of novel models for handling imbalanced datasets.

The challenge to automatically select the best among models of varying dimensions remains open, especially in the context of complex models, sparse data, and noisy data. Bayesian neural networks employ Markov chain Monte Carlo (MCMC) and variational inference methods for training (sampling) model parameters. However, the progress of MCMC methods in deep learning has been slow due to high computational requirements and uninformative priors of model parameters. Reversible jump MCMC allows sampling of model parameters of variable lengths; hence, it has the potential to train Bayesian neural networks effectively. In this paper, we implement reversible jump MCMC for training dynamic Bayesian neural networks that feature cascaded neural networks with dynamic hidden and input neurons. We apply the methodology to a wide range of regression and classification problems from the literature. The results show that our proposed framework provides an effective approach for the dynamic exploration of models while featuring uncertainty quantification that not only caters to model parameters but also extends to model topology. This opens up the road for uncertainty quantification in dynamic neural networks where hidden and input neurons can change over time.

Bayesian neural networks harness the power of Bayesian inference which provides an approach to neural learning that not only focuses on accuracy, but also uncertainty quantification. Markov Chain Monte Carlo (MCMC) methods implement Bayesian inference by sampling from the posterior distribution of the model parameters. In the case of Bayesian neural networks, the model parameters refer to weights and biases. MCMC methods suffer from scalability issues in large models, such as deep neural networks with thousands to millions of parameters. In this paper, we present a Bayesian ensemble learning framework that utilises gradient boosting by combining multiple shallow neural networks (base learners) that are trained by MCMC sampling. We present two Bayesian gradient boosting strategies that employ simple neural networks as base learners with Langevin MCMC sampling. We evaluate the performance of these methods on various classification and time-series prediction problems. We demonstrate that the proposed framework improves prediction accuracy of canonical gradient boosting, while providing uncertainty quantification via Bayesian inference. Furthermore, we demonstrate that the respective methods scale well when the size of the dataset and model increases.

It is well known that canonical recurrent neural networks (RNNs) faced limitations in learning long-term dependencies which has been addressed by memory structures in long short-term memory (LSTM) networks. Neural Turing machines (NTMs) are novel RNNs that implement the notion of programmable computers with neural network controllers which can learn simple algorithmic tasks. Matrix neural networks feature matrix representation which inherently preserves the spatial structure of data when compared to canonical neural networks that use vector-based representation. The matrix-representation of neural networks also have the potential to provide better memory capacity. In this paper, we define and study a probabilistic notion of memory capacity based on Fisher information for matrix-based RNNs. We find bounds on memory capacity for such networks under various hypotheses and compare them with their vector counterparts. In particular, we show that the memory capacity of such networks is bounded by N2 for N×N state matrix which generalizes the one known for vector networks. We also show and analyze the increase in memory capacity for such networks which is introduced when one exhibits an external state memory, such as Neural Turing Machines (NTMs). This motivates us to construct NTMs with RNN controllers with matrix-based representation of external memory, leading us to introduce Matrix NTMs. We demonstrate the performance of this class of memory networks under certain algorithmic learning tasks such as copying and recall and compare it with Matrix RNNs. We find an improvement in the performance of Matrix NTMs by the addition of external memory.

In certain types of problems, such as emerging storms or cyclones, robust prediction is needed even when partial information is available. Dynamic time series prediction refers to “on the fly” prediction given partial information. Recently, a neu-roevolution approach called co-evolutionary multi-task learning has been proposed to provide robust prediction for dynamic time series. In this paper, we adapt the method with multi-task modular backpropagation that features gradient descent and transfer learning. The method is tested on benchmark chaotic time series problems and compared with its counterparts. The results show that the method can alleviate the problems associated with timely convergence of the neuroevolution approach and provides better performance.

Time series prediction typically consists of a data reconstruction phase where the time series is broken into overlapping windows. The size of the window could vary for different types of problems for optimal performance. Dynamic time series prediction refers to “on the fly'' robust prediction given partial information where prediction can be made regardless of the window size. Multi-task learning features learning from related tasks through shared representation knowledge which has shown to be useful for dynamic time series prediction. This features uncertainty that can be addressed through synergy of Bayesian inference and multi-task learning. In this paper, we present a Bayesian approach to multi-task learning for dynamic time series prediction. The method provides uncertainty quantification given posterior distribution of weights and biases in a cascaded multitask network architecture. The results show that the proposed method is able to provide competing prediction performance to the literature, featuring uncertainty quantification in prediction.

Memetic algorithms have been a promising strategy to enhance neuroevolution in the past. Cooperative coevolution has been combined as memetic cooperative neuroevolution with application to chaotic time series prediction. Although the method has shown promising performance, there are limitations in the balance between global and local search. The previous study used a specific local search strategy for intensification that affected the diversity of solutions. In this study, we address this limitation by information (meme) collection strategies that maintains and refines a pool of memes during global search. We present two strategies where one is sequential and the other is concurrent meme collection implemented at different stages of evolution. In the majority of the given problems, the proposed strategies showed improvement in prediction accuracy over the related methods.

Although machine learning and statistical methods have been extensively used to study cyclones, the prediction of cyclone trajectories remains a challenging problem. Matrix neural networks have the ability of handling spatial correlations in the data which made them suitable for image recognition tasks. Cyclone trajectories are defined by the latitude and the longitudes coordinates as a temporal sequence. Matrix neural networks are suitable for track prediction as the dataset can be conveniently given as input without vectorization that could result in loss of correlation of the spatial information. In this paper, matrix neural networks are used for cyclone track prediction for the South Indian Ocean. The results show that matrix neural networks have the ability to preserve spatial correlation that empower them to make a better prediction when compared to prominent recurrent neural network architectures.

Competition in cooperative coevolution (CC) has demonstrated success in solving global optimization problems. In a recent study, a multi-island competitive cooperative coevolution (MIC3) algorithm was introduced that featured competition and collaboration of several different problem decomposition strategies implemented as independent islands. It was shown that MIC3converges to high quality solutions without the need to find an optimal decomposition. MIC3splits the computational budget in terms of the number of function evaluations, equally amongst all the islands and evolves them in a round-robin fashion. This overlooks the difference in contributions of different islands towards improving the overall objective function value. Therefore, a considerable amount of function evaluations is wasted on the low-contributing islands as their problem decomposition strategies may not appeal to the problem at the given stage of the evolutionary process. This paper proposes contribution-based MIC3algorithms (MIC4) that quantifies the contributions of each island and allocates the computational budget accordingly. The experimental analysis reveals that the proposed method outperforms its counterpart.

Problem decomposition determines how subcomponents are created that have a vital role in the performance of cooperative coevolution. Cooperative coevolution naturally appeals to fully separable problems that have low interaction amongst subcomponents. The interaction amongst subcomponents is defined by the degree of separability. Typically, in cooperative coevolution, each subcomponent is implemented as a sub-population that is evolved in a round-robin fashion for a specified depth of evolution. This paper examines the relationship between the depth of evolution and degree of separability for different types of global optimisation problems. The results show that the depth of evolution is an important attribute that affects the performance of cooperative coevolution and can be used to ascertain the nature of the problem in terms of the degree of separability.

Multi-step-ahead time series prediction has been one of the greatest challenges for machine learning. Recurrent neural networks (RNN) can efficiently model temporal sequences and have been promising for multi-step time series prediction. Cooperative neuro-evolution has been used for training RNNs with promising performance for single step ahead time series prediction. This paper employs cooperative neuro-evolution of RNNs for multi-step ahead prediction. The RNN recursively predicts the next values in the horizon where the output from the single-step ahead prediction are the input for predicting the next value in the horizon. The performance of cooperative neuro-evolution is compared with back-propagation through time (BPTT) learning algorithm. The results are promising which shows that cooperative neuro-evolution performs better compared to BPTT for most cases.

Reliable prediction of solar power is very important to increase the penetration of this renewable energy source into the electricity grid. In this paper, we consider the task of forecasting solar power output from Photovoltaic (PV) systems at half-hourly intervals. We propose a new approach based on recurrent neural networks trained with cooperative neuro-evolution algorithm. We develop both univariate model which uses only previous power data and multivariate model which uses both previous power and weather data for prediction. We conduct a comprehensive evaluation of the proposed approach using 2 years of solar power data from a grid-connected PV plant. Evaluation shows that proposed approach achieves promising accuracy outperforming the persistence models used as baselines.

Cyclone track prediction is a two dimensional time series prediction problem that involves latitudes and longitudes which define the position of a cyclone. Recurrent neural networks have been suitable for time series prediction due to their architectural properties in modeling temporal sequences. Coevolutionary recurrent neural networks have been used for time series prediction and also applied to cyclone track prediction. In this paper, we present an architecture for encoding two dimensional time series problem into Elman recurrent neural networks composed of a single input neuron. We use cooperative coevolution and back-propagation through-time algorithms for training. Our experiments show an improvement in the accuracy when compared to previous results using a different recurrent network architecture.

Time series prediction relies on past data points to make robust predictions. The span of past data points is important for some applications since prediction will not be possible unless the minimal timespan of the data points is available. This is a problem for cyclone wind-intensity prediction, where prediction needs to be made as a cyclone is identified. This paper presents an empirical study on minimal timespan required for robust prediction using Elman recurrent neural networks. Two different training methods are evaluated for training Elman recurrent network that includes cooperative coevolution and backpropagation-though time. They are applied to the prediction of the wind intensity in cyclones that took place in the South Pacific over past few decades. The results show that a minimal timespan is an important factor that leads to the measure of robustness in prediction performance and strategies should be taken in cases when the minimal timespan is needed.

Cooperative coevolution is an evolutionary computation method which solves a problem by decomposing it into smaller subcomponents. Multi-objective optimization deals with conflicting objectives and produces multiple optimal solutions instead of a single global optimal solution. In previous work, a multi-objective cooperative co-evolutionary method was introduced for training feedforward neural networks on time series problems. In this paper, the same method is used for training recurrent neural networks. The proposed approach is tested on time series problems in which the different time-lags represent the different objectives. Multiple pre-processed datasets distinguished by their time-lags are used for training and testing. This results in the discovery of a single neural network that can correctly give predictions for data pre-processed using different time-lags. The method is tested on several benchmark time series problems on which it gives a competitive performance in comparison to the methods in the literature.

Cooperative coevolution has proven to be efficient in solving global optimisation and real world application problems. However, it is highly sensitive to problem decomposition, especially in the context of non-separable functions that possess interacting decision variables. Problem decomposition has been a challenge of cooperative coevolution. Efficient problem decomposition strategy ensures that interacting variables are grouped into separate subcomponents. Introduction of competition and collaboration features have shown to be advantageous in evolutionary algorithms but have not quite been fully explored in cooperative coevolution. In this paper, a method is utilized that enforces competition in coevolution whereby different problem decomposition schemes are implemented as islands that compete and collaborate with each other. The proposed framework is tested on several global optimisation benchmark problems and achieves promising results.

Climate change issues are continuously on the rise and the need to build models and software systems for management of natural disasters such as cyclones is increasing. Cyclone wind-intensity prediction looks into efficient models to forecast the wind-intensification in tropical cyclones which can be used as a means of taking precautionary measures. If the wind-intensity is determined with high precision a few hours prior, evacuation and further precautionary measures can take place. Neural networks have become popular as efficient tools for forecasting. Recent work in neuro-evolution of Elman recurrent neural network showed promising performance for benchmark problems. This paper employs Cooperative Coevolution method for training Elman recurrent neural networks for Cyclone wind-intensity prediction in the South Pacific region. The results show very promising performance in terms of prediction using different parameters in time series data reconstruction.

This paper presents a two-dimensional time series prediction approach for cyclone track prediction using cooperative neuro-evolution of Elman recurrent networks in the South Pacific region. The latitude and longitude of tracks of cyclone lifetime is taken into consideration for past three decades to build a robust forecasting system. The proposed method performs one step ahead prediction of the cyclone position which is essentially a two-dimensional time series prediction problem. The results show that the Elman recurrent network is able to achieve very good accuracy in terms of prediction of the tracks which can be used as means of taking precautionary measures.

In the application of cooperative coevolution for neuro-evolution, problem decomposition methods rely on architectural properties of the neural network to divide it into subcomponents. During every stage of the evolutionary process, different problem decomposition methods yield unique characteristics that may be useful in an environment that enables solution sharing. In this paper, we implement a two-island competition environment in cooperative coevolution based neuro-evolution for feedforward neural networks for pattern classification problems. In particular the combinations of three problem decomposition methods that are based on the architectural properties that refers to neural level, network level and layer level decomposition. The experimental results show that the performance of the competition method is better than that of the standalone problem decomposition cooperative neuro-evolution methods.

Memetic algorithms (MA) are evolutionary computation methods that employ local search to selected individuals of the population. This work presents global–local population MA for solving the forward kinematics of parallel manipulators. A real-coded generation algorithm with features of diversity is used in the global population and an evolutionary algorithm with parent-centric crossover operator which has local search features is used in the local population. The forward kinematics of the 3RPR and 6–6 leg manipulators are examined to test the performance of the proposed method. The results show that the proposed method improves the performance of the real-coded genetic algorithm and can obtain high-quality solutions similar to the previous methods for the 6–6 leg manipulator. The accuracy of the solutions and the optimisation time achieved by the methods in this work motivates for real-time implementation of the 3RPR parallel manipulator.

Collaboration enables weak species to survive in an environment where different species compete for limited resources. Cooperative coevolution (CC) is a nature-inspired optimization method that divides a problem into subcomponents and evolves them while genetically isolating them. Problem decomposition is an important aspect in using CC for neuroevolution. CC employs different problem decomposition methods to decompose the neural network training problem into subcomponents. Different problem decomposition methods have features that are helpful at different stages in the evolutionary process. Adaptation, collaboration, and competition are needed for CC, as multiple subpopulations are used to represent the problem. It is important to add collaboration and competition in CC. This paper presents a competitive CC method for training recurrent neural networks for chaotic time-series prediction. Two different instances of the competitive method are proposed that employs different problem decomposition methods to enforce island-based competition. The results show improvement in the performance of the proposed methods in most cases when compared with standalone CC and other methods from the literature.

The use of neural networks for time series prediction has been an important focus of recent research. Multi-objective optimization techniques have been used for training neural networks for time series prediction. Cooperative coevolution is an evolutionary computation method that decomposes the problem into subcomponents and has shown promising results for training neural networks. This paper presents a multi-objective cooperative coevolutionary method for training neural networks where the training data set is processed to obtain the different objectives for multi-objective evolutionary training of the neural network. We use different time lags as multi-objective criterion. The trained multi-objective neural network can give prediction of the original time series for preprocessed data sets distinguished by their time lags. The proposed method is able to outperform the conventional cooperative coevolutionary methods for training neural networks and also other methods from the literature on benchmark problems.

Intelligent financial prediction systems guide investors in making good investments. Investors are continuously on the hunt for better financial prediction systems. Neural networks have shown good results in the area of financial prediction. Cooperative coevolution is an evolutionary computation method that decomposes the problem into subcomponents and has shown promising results for training neural networks. This paper presents a computational intelligence framework for financial prediction where cooperative coevolutionary feedforward neural networks are used for predicting closing market prices for companies listed on the NASDAQ stock exchange. Problem decomposition is an important step in cooperative co-evolution that affects its performance. Synapse and Neuron level are the main problem decomposition methods in cooperative coevolution. These two methods are used for training neural networks on the given financial prediction problem. The results show that Neuron level problem decomposition gives better performance in general. A prototype of a mobile application is also given for investors that can be used on their Android devices.

Problem decomposition is an important aspect in using cooperative coevolution for neuro-evolution. Cooperative coevolution employs different problem decomposition methods to decompose the neural network training problem into subcomponents. Different problem decomposition methods have features that are helpful at different stages in the evolutionary process. Adaptation, collaboration and competition are characteristics that are needed for cooperative coevolution as multiple sub-populations are used to represent the problem. It is important to add collaboration and competition in cooperative coevolution. This paper presents a competitive two-island cooperative coevolution method for training recurrent neural networks on chaotic time series problems. Neural level and Synapse level problem decomposition is used in each of the islands. The results show improvement in performance when compared to standalone cooperative coevolution and other methods from literature.

Cooperative coevolution employs different problem decomposition methods to decompose the neural network problem into subcomponents. The efficiency of a problem decomposition method is dependent on the neural network architecture and the nature of the training problem. The adaptation of problem decomposition methods has been recently proposed which showed that different problem decomposition methods are needed at different phases in the evolutionary process. This paper employs an adaptive cooperative coevolution problem decomposition framework for training recurrent neural networks on chaotic time series problems. The Mackey Glass, Lorenz and Sunspot chaotic time series are used. The results show improvement in performance in most cases, however, there are some limitations when compared to cooperative coevolution and other methods from literature.

Memetic algorithms and cooperative coevolution are emerging fields in evolutionary computation which have shown to be powerful tools for real-world application problems and for training neural networks. Cooperative coevolution decomposes a problem into subcomponents that evolve independently. Memetic algorithms provides further enhancement to evolutionary algorithms with local refinement. The use of crossover-based local refinement has gained attention in memetic computing. This paper employs a cooperative coevolutionary framework that utilises the strength of local refinement via crossover. The framework is evaluated by training recurrent neural networks on grammatical inference problems. The results show that the proposed approach can achieve better performance than the standard cooperative coevolution framework.

In this paper, an adaptive modularity cooperative coevolutionary framework is presented for training feedforward neural networks. The modularity adaptation framework is composed of different neural network encoding schemes which transform from one level to another based on the network error. The proposed framework is compared with canonical cooperative coevolutionary methods. The results show that the proposal outperforms its counterparts in terms of training time, success rate and scalability.