Recent game content generation using artificial intelligence (AI) has been limited by a lack of diversity in content styles, and the impact of human input on generated content has not received enough attention. To address these, we present an innovative game content generation framework for enhancing the game style of Super Mario Bros. This framework leverages the dual advantages of interactive evolutionary computation (IEC) for capturing players' implementation preferences and style-based generative adversarial network (StyleGAN) for generating high-quality and diverse solutions. Additionally, we introduce a novel human-computer interaction (HCI) console that allows users to explore the generative adversarial network's latent space, guiding content generation towards their preferences. Finally, we analyze the performance of the proposed framework from both quantitative and qualitative perspectives. The results demonstrate that the framework model significantly enhances the diversity of the generated content. Moreover, it sheds light on the role of human participation in the content creation process.

Deep learning performs feature extraction through a series of data transformations. Convolutional neural networks (CNNs) are among the most representative methods in deep learning. CNNs enable complex data transformations through convolution operations to extract high-level features. The kernel-based autoencoder (KAE) uses kernelized principal component analysis to perform both linear and nonlinear data transformations. Consequently, KAE provides better interpretability than CNNs in feature extraction. This paper presents a comprehensive study of KAE, including both single-layer and multi-layer configurations. Two distinct multi-layer autoencoder architectures based on the kernel method are proposed. The proposed method is evaluated using handwritten digit images. Experimental results demonstrate the proposed method’s effectiveness in feature extraction. In terms of the structural similarity index measure, the multi-layer kernel-based autoencoder achieves results comparable to those of the convolutional autoencoder. With fewer training samples, the multi-layer kernel-based autoencoder outperforms the convolutional autoencoder.

Accurate prediction of intraoperative hazardous events and generation of effective intervention plans are critical to surgical safety, but face multiple challenges of real-time, accuracy, and interpretability. Large-scale language models have potential, but their high cost and potential ‘illusion’ problems limit their application in real-time clinical environments. Traditional multitask learning models are efficient but knowledge-constrained, making it difficult to capture complex reasoning processes. To bridge this gap, this paper proposes a multi-objective distillation knowledge enhancement model-KEMO, which innovatively adopts a multi-objective chain-of-thought distillation framework to not only mimic the prediction results of the instructor’s LLM, but also explicitly migrate its structured reasoning process to the lightweight student model, which improves the answerability of the model by synergistically optimising the three objectives of event prediction, reasoning alignment and scenario generation. Interpretability. Meanwhile, combined with the Knowledge Graph-based Retrieval Augmented Generation mechanism, validated medical knowledge is dynamically injected to enhance the accuracy and reliability of decision-making and reduce model illusion. The experimental results show that the KEMO model significantly outperforms traditional models of the same magnitude in intraoperative hazardous event prediction and prognostic proposal generation, and achieves a performance comparable to that of a large faculty model.The KEMO model effectively bridges the gap between the large language model and the actual clinical application, and facilitates the transformation of the large model knowledge to the actual clinical deployment.

In evolutionary computation, the study subject for algorithms capable of effectively resolving multi-objective optimization problems remains at the forefront of research. Most existing multi-objective evolutionary algorithms (MOEAs) often struggle with maintaining diversity and avoiding premature convergence when solving complex or high-dimensional optimization problems. This study proposes an innovative iteration of the Non-dominated Sorting Genetic Algorithm III (NSGA-III), which infuses chaotic dynamics alongside estimated convergence point strategies to enhance solution quality and diversity. Our research undertakes a comprehensive evaluation of this enhanced algorithm against a suite of benchmark problems. We compare it with its predecessor, Non-dominated Sorting Genetic Algorithm II (NSGA-II), and variations incorporating chaotic dynamics alongside estimated convergence point strategies. The performance analysis results indicate that NSGA-III, using chaotic dynamics and estimated convergence point strategies across most test functions, demonstrates superior performance over traditional and singly-enhanced MOEAs. The statistical analysis strongly suggests that the dual enhancements embedded in the novel algorithm contribute significantly to its optimization capabilities.

Balancing exploration and exploitation is a fundamental challenge in reinforcement learning. In Maximum Diffusion Reinforcement Learning (MaxDiff RL), this balance is regulated by a temperature parameter that controls exploration, but optimizing it across different tasks is challenging. To tackle this issue, we propose a method that dynamically adjusts the temperature parameter via evolutionary algorithms during training. The training process is divided into multiple stages, where different optimization strategies are applied to adaptively evolve the temperature parameter, ensuring a proper balance between exploration and exploitation. We evaluate our method on two continuous control tasks in robotics, i.e., Swimmer and HalfCheetah. Experimental results demonstrate that our method outperforms baseline algorithms with randomly initialized or default temperature parameters, achieving faster convergence and higher cumulative rewards, particularly in tasks demanding greater exploration.

One-shot neural architecture search (NAS) trains an over-parameterized network (termed as supernet) that assembles all the architectures as its subnets by using weight sharing for computational budget reduction. However, there is an issue of multi-model forgetting during supernet training that some weights of the previously well-trained architecture will be overwritten by that of the newly sampled architecture which has overlapped structures with the old one. To overcome the issue, we propose an orthogonal gradient learning (OGL) guided supernet training paradigm, where the novelty lies in the fact that the weights of the overlapped structures of current architecture are updated in the orthogonal direction to the gradient space of these overlapped structures of all previously trained architectures. Moreover, a new approach of calculating the projection is designed to effectively find the base vectors of the gradient space to acquire the orthogonal direction. We have theoretically and experimentally proved the effectiveness of the proposed paradigm in overcoming the multi-model forgetting. Besides, we apply the proposed paradigm to two one-shot NAS baselines, and experimental results demonstrate that our approach is able to mitigate the multi-model forgetting and enhance the predictive ability of the supernet with remarkable efficiency on popular test datasets.

This paper presents an innovative approach employing persistence-based clustering in Riemannian manifolds within evolutionary computation algorithms to address multi-modal optimization problems. The proposed framework is im-plemented and evaluated using the chaotic evolution algorithm. We introduce a novel algorithm named chaotic evolution with a clustering algorithm (CECA), which integrates the chaotic evolution characteristics from chaotic systems with the clustering method and Gaussian local search to solve multi-modal optimization problems. By leveraging chaotic dynamics, CECA enhances exploration and exploitation for efficient searching. Simultane-ously, it utilizes the clustering method to improve population diversity in the context of multi-modal optimization problems. The effectiveness and advantages of the proposed framework on the CECA algorithm are demonstrated through extensive experimental evaluations of various benchmark functions, in-cluding the Congress on Evolutionary Computation (CEC) con-ference functions. The experimental results indicate that the proposed framework exhibits distinct advantages in optimizing high-dimensional complex multi-modal functions. This study provides empirical evidence that persistence-based clustering in Riemannian manifolds constitutes an effective methodology for evolutionary multi-modal optimization.

In tumor diagnosis, due to subtle differences in the imaging appearance of different diseases, accurately classifying lesions based on solely imaging data proves challenging. Existing machine learning and deep learning methods face limitations due to the small sample size of medical datasets and the intricate nature of disease image manifestations. This paper proposes a novel lesion classification method to fully explore distinctions among confused lesion features associated with different diseases. The proposed method comprises three key steps: Firstly, a lesion feature calculation method using dynamic fusion of global attention and local attention is proposed. The weight of global attention and local attention is dynamically allocated, and the global and local features are fused by dynamic weight. Secondly, feature dimension reduction is realized to improve the effect of distinguishable features using sparse autoencoder and polynomial constraint loss function. Finally, to improve the performance of classification, the monarch butterfly optimization algorithm based on adaptive neighborhood search radius method is used to optimize the parameters of multi-kernel support vector machine. The private PET/CT image classification dataset of lymphoma and Still’s disease was used to validate our results. The experimental results demonstrate that the method's efficacy in lymphoma and Still’ disease classification tasks, achieving an accuracy (ACC) of 82.8% and an area under the curve (AUC) of 87.1%, respectively.

We propose an extension of the chaotic evolution algorithm into the discrete domain to address combinatorial optimization problems. In this study, we leverage the discrete chaotic evolution algorithm to tackle the Traveling Salesman Problem (TSP) for assessment purposes. The chaotic evolution algorithm exploits the ergodicity of chaos to facilitate the search process within the optimization algorithm. It incorporates a mathematical mechanism into the iterative evolution process, simulating ergodic motion within a search space based on a simple principle. To manage the discrete mutation operation within the chaotic evolution algorithm, we introduce a specifically designed chaotic operation. This operation is tailored for its application in solving combinatorial optimization problems. The chaotic sequence plays a crucial role in determining the mutation location. Our evaluation involves the comparison of our proposed discrete chaotic evolution algorithm with the outcomes of the simulated annealing algorithm and the tabu search algorithm. The assessment serves to demonstrate and validate that the discrete chaotic evolution algorithm yields superior optimization performance within the discrete domain.

Accurately estimating the temperature and salinity structure of lakes or reservoirs is crucial for understanding terrestrial hydrological processes and pollutant transport path-ways. However, key parameters for solute transport models in hydrodynamic systems are difficult to obtain directly and often require inversion simulations involving multi-source solute parameters. This study addresses the challenges of multi-source heterogeneous data assimilation and the optimization of training data distribution for alternative models. We develop a hybrid algorithm that solves data assimilation issues for multi-source heterogeneous data in reactive solute transport inversion simulations. By combining posterior inference of states with the prior distribution of parameters, we propose a novel coupled paradigm and explore the impact of identifying characteristic parameters across multiple scenarios and the performance advantages of surrogate models.

The simulation and analysis of complex spatiotemporal systems are crucial for expressing and solving chaotic dynamical systems such as those in Earth and environmental sciences. Understanding and computing physical processes, re-actions, or substance transport typically relies on control equations. This paper aims to explore a novel research paradigm by enhancing the physical network coupling structure to construct predictive models for fluid dynamics systems, simulating spatiotemporal dynamical processes of substance transformation in the domain of environmental physics. In particular, when addressing problems involving the non-homogeneous 2D fluid dynamics equations, the characteristic parameters of the phys-ical processes were redefined. This was achieved by encoding hard boundary conditions and designing appropriate neural network architectures to mitigate over-fitting issues during the prediction of parameterized dynamical systems. Comparative experiments involving five benchmark physics-informed neural network methods emphasize the significant improvement in capturing time-varying features and prediction accuracy brought by the proposed approach. Through various water cycle scenarios, the model's estimation ability for diffusion fields is validated, focusing on analyzing the influence of data errors and sample size on the computational results of this deep neural network. Notably, the proposed method exhibits higher robustness to outlier observations under extreme conditions.

We propose a data analysis method that combines the objectives of nonlinear principal component analysis and nonlinear discriminant analysis with the kernel method in a reproducing kernel Hilbert space. This method addresses nonlinear data analysis problems in high-dimensional spaces, specifically the reproducing kernel Hilbert space, through the use of the kernel trick. Our proposed method can be considered as a semi-supervised data analysis approach. We evaluate our proposed method using various kernel functions and datasets, both visually and quantitatively. The evaluation results demonstrate that our proposal outperforms kernel principal component analysis and generalized discriminant analysis in terms of classification performance. This indicates the advantages and originality of our proposed method. Furthermore, we analyze and discuss our findings based on the evaluation results, and highlight potential areas for further research and future work related to our proposal.

Over the past several years, deep learning technologies have made tremendous progress in medical image tasks including classification, segmentation, and object detection. However, there are two main limitations of indocyanine green (ICG) images which are often used in breast cancer related lymphedema (BCRL): insufficient sample numbers and low image quality. Consequently, the conventional deep learning based classification methods such as ResNet have faced challenges in achieving satisfactory results. To tackle the concern, this paper puts forward a deep learning method named Denoising Diffusion Probabilistic Model Self-supervised Knowledge Distillation Net (DDPM-SKDNet) for the ICG images classification task, by incorporating a contrastive learning based approach as the network architecture and using DDPM as the image generator in the contrastive module to expand the dataset size. Furthermore, a knowledge distillation approach is utilized to increase the effectiveness of the network. The proposed method was validated on ICG datasets and achieved a significant improvement in classification accuracy, increasing it from 66.7% in the baseline method to 82.1% in the proposed method.

We propose a method that combines multi-dimensional scaling (MDS) with a regression model to approximate fitness landscapes and enhance the efficiency of evolutionary search algorithms. The approach involves projecting the original population onto a low-dimensional space using MDS, thereby preserving pairwise distances between data points. The optimal individual is determined using the approximated fitness landscape in the low-dimensional space and subsequently mapped back to the original space using our designed rules. Furthermore, several prediction points are generated in the proximity of the optimal individual to replace certain individuals. This iterative process is repeated multiple times to accelerate the search speed of evolutionary computation (EC). Experimental results demonstrate the superiority of our proposed method over the other two algorithms across most tested functions.

There exists an inherent contradiction between the continuous accumulation of environmental-geographical spatiotemporal data and the ongoing refinement of analysis granularity. Algorithms aimed at mining spatiotemporal water quality data in river networks are confronted with constraints posed by statistical factors, including sparse distribution, heterogeneity, and spatiotemporal autocorrelation features. Consequently, prevailing machine learning methods encounter challenges when attempting to encapsulate the evolving characteristics of non-stationarity over time and the cyclic evolution patterns of intricate watershed geographical processes. These challenges serve as notable impediments to achieving precise and resilient predictions for watershed water quality. In consideration of these circumstances, we propose a predictive framework built upon the fusion of deep learning and multi-view semi-supervised learning. This framework involves the modeling of correlations between time series feature factors and water quality, the discernment of relationships between the spatial-physical structure and temporal features of river networks, and the identification of similar features related to the environmental context across distinct river segments. Additionally, we introduce coefficients for adjusting feature weights to mitigate the adverse impact stemming from noise or data sparsity in low-quality views. We establish a feature classifier that upholds inter-view consistency and global spatiotemporal correlations, leveraging complementary knowledge from different feature perspectives to capture long-range spatiotemporal dependencies among river segments. This, in turn, enhances cross-watershed global prediction accuracy. Through comprehensive performance comparisons with eight competing algorithms, the results underscore that the proposed approach demonstrates enhanced robustness and superiority, particularly within intricate and complex river network systems.

The effective classification of pollen is a critical method to the prevention of pollen allergy. The conventional pollen classification primarily relies on manual handling under the microscope, which does not only require a lot of manpower but also has low classicization accuracy on results. In this paper, the optical microscope was used to scan the slides and the pollen image dataset was made for classification. We found from the pollen dataset that most of the images have clear contours, but there were also many pollen images with the following problems. One is that the pollen was covered by impurities like dust and pebbles, and the other is that the pollen images stain unevenly and the pollen image was blurred and the number of pollens was unbalanced. This paper proposes a Dual-Channel Attention method, we call it DCANet, which combines the advantages of channel attention and channel self-attention to improve the classification accuracy. The results of classification activation map analysis showed that DCANet paid more attention to pollen images.

This paper proposes a novel population-based optimization algorithm to solve the multi-modal optimization problem. We call it the chaotic evolution deterministic crowding (CEDC) algorithm. Since the genetic algorithm is difficult to find all optimal solutions and the accuracy is not high when searching for multi-modal optimization problems, we use the ergodicity of chaos to implement the exploration and fitness comparison of the deterministic crowding algorithm. Through the tests of several multi-modal benchmark functions, it is shown that the algorithm can effectively and accurately find the most optimal solutions to the multi-modal problem. It does not need to set the niche radius in advance, so it can better solve multimodal optimization problems. We test it with nine multi-modal benchmark functions ranging from one-dimension (1-D) to ten-dimension (10-D), and we compare it with a genetic algorithm and evaluate from peak ratio, max peak ratio, and running time. The experimental results show that the CEDC algorithm is better than conventional algorithms in both runtime and peak ratio.

We introduce two new strategies into the conventional chaotic evolution (CE) to facilitate the exchange of information between individuals instead of searching independently, and propose a cooperative CE (CoCE) with stronger performance. The first strategy designs a new crossover operation to ensure that mutant individual's exchange genes with other better individuals. Specifically, a mutant individual does not cross with its parent individual, but with a randomly selected individual whose fitness is not worse than that of the parent individual. The second strategy introduces an additional attraction generated by the randomly selected individual acting on the mutant individual, which can allow the offspring individual to prefer potential areas and speed up the search process. To evaluate the performance of our proposed strategies, we configure CoCE and CE with the exact same parameter settings and run them on 28 benchmark functions from CEC 2013 test suites. Each benchmark function is run 30 times independently on three different dimensions (i.e., 2-D, 10-D, and 30-D), and we also apply the Friedman test and Holm's multiple comparison test to check significant differences between CE and its three variants at the termination condition. The experimental results confirmed that our proposed CoCE has a faster convergence speed and higher convergence accuracy, and the acceleration effect is more significant for high dimensional optimization problems.

Several developments in computational image processing methods assist the radiologist in detecting abnormal breast tissue in recent years. Consequently, deep learning-based models have become crucial for early screening and interpretation of mammographic images for breast masses diagnosis, helping for successful treatment. Breast masses and calcification is an essential parameter for the prognosis of breast cancer. However, the mammographic image’s mass detection needs a deeper investigation due to the breast masses’ heterogeneity and anomalies’ characteristics that are easily confused with other objects present in the image. Hence, this study proposed a deep learning-based convolutional neural network (ConvNet) that will incorporate both mammography and clinical variables to predict and classify breast masses to assist the expert’s decision-making processes. We trained our proposed model with 322 scanned digital mammographic images of the MIAS (Mammogram Image Analysis Society) dataset and 580 images of the private dataset to evaluate the performance, which is highly imbalanced. This study aimed to perform an automatic and comprehensive characterization of breast masses using appropriate layers deep ConvNet model with high accuracy true-positive rate, decreased error rate and applying data-augmentation techniques. We obtained a classification accuracy of 97% applying the filtered deep features, which is the best performance from the existing approaches.

Chaotic Evolution (CE) is simply a population-based algorithm that applies the ergodic properties of the chaotic system into the search process. With the help of the chaotic motion, the algorithm can visit any arbitrary point with a movement track. However, its search performance should be enhanced to get a better algorithm as CE spends a lot of invalid computation costs when solving the complex optimization problem due to premature convergence and/or slow convergence. In this paper, we introduce, make a comparison, and evaluate the effectiveness of the proposed CE with Elite strategy (ECE), which is based on the heuristic information from the elite individuals of the current population. Moreover, because in CE, the distribution characteristic of the chaotic system plays a very important role in deciding the search power, we also make paired comparison and analyze the optimization performance between our proposed ECE algorithms using different types of chaotic map and the canonical CE by applying 12 well-known Benchmark functions as the test functions and some measurement tests. The investigation process indicates that the ECE algorithm is significantly better on the majority of the test functions.

Large for gestational (LGA) means the fetus having an abnormal birth weight. It adheres severe complications during and after the maternal period. Therefore, this research presents an ensemble classification scheme using Chinese National Pre-Pregnancy Examination Program dataset to classify a fetus as an LGA or non-LGA based on provided Chinese LGA classification guidelines. Moreover, the proposed scheme is comprised of data cleansing and ensemble classification schemes that have drastically improved the LGA classification process with improved performance results compared to present published studies. Therefore, the recommended scheme can be utilized by healthcare professionals to build an enhanced and reliable LGA classification system.

Recently, the automatic diagnosis of Turner syndrome (TS) has been paid more attention. However, existing methods relied on handcrafted image features. Therefore, we propose a TS classification method using unsupervised feature learning. Specifically, first, the TS facial images are preprocessed including aligning faces, facial area recognition and processing of image intensities. Second, pre-trained convolution filters are obtained by K-means based on image patches from TS facial images, which are used in a convolutional neural network (CNN); then, multiple recursive neural networks are applied to process the feature maps from the CNN to generate image features. Finally, with the extracted features, support vector machine is trained to classify TS facial images. The results demonstrate the proposed method is more effective for the classification of TS facial images, which achieves the highest accuracy of 84.95%.

We propose a chaotic evolution algorithm with elite strategy. The conventional chaotic evolution algorithm uses each individual to search in its local area. The proposed algorithm searches the parameter space always around the elite individual from the last generation. We evaluate the proposed algorithm in both single-objective and multi-objective optimization problems. In the single objective optimization problem, the elite is the individual has the best fitness value, and in the multi-objective optimization problem, the elites are the individuals in the first Pareto front. We design and evaluate these two algorithms with elite strategy using single- and multi-objective benchmark functions. We design a jump strategy to avoid searching within a local optima areas by applying elite strategy several generations one time. The numerical evaluation results demonstrate the proposed algorithm has strong local exploitation capability in the early generations. The optimization performance of chaotic evolution algorithm has a potential possibility to apply in high dimensional and more complex optimization problems.

BACKGROUND: Screening of the brain computerised tomography (CT) images is a primary method currently used for initial detection of patients with brain trauma or other conditions. In recent years, deep learning technique has shown remarkable advantages in the clinical practice. Researchers have attempted to use deep learning methods to detect brain diseases from CT images. Methods often used to detect diseases choose images with visible lesions from full-slice brain CT scans, which need to be labelled by doctors. This is an inaccurate method because doctors detect brain disease from a full sequence scan of CT images and one patient may have multiple concurrent conditions in practice. The method cannot take into account the dependencies between the slices and the causal relationships among various brain diseases. Moreover, labelling images slice by slice spends much time and expense. Detecting multiple diseases from full slice brain CT images is, therefore, an important research subject with practical implications. RESULTS: In this paper, we propose a model called the slice dependencies learning model (SDLM). It learns image features from a series of variable length brain CT images and slice dependencies between different slices in a set of images to predict abnormalities. The model is necessary to only label the disease reflected in the full-slice brain scan. We use the CQ500 dataset to evaluate our proposed model, which contains 1194 full sets of CT scans from a total of 491 subjects. Each set of data from one subject contains scans with one to eight different slice thicknesses and various diseases that are captured in a range of 30 to 396 slices in a set. The evaluation results present that the precision is 67.57%, the recall is 61.04%, the F1 score is 0.6412, and the areas under the receiver operating characteristic curves (AUCs) is 0.8934. CONCLUSION: The proposed model is a new architecture that uses a full-slice brain CT scan for multi-label classification, unlike the traditional methods which only classify the brain images at the slice level. It has great potential for application to multi-label detection problems, especially with regard to the brain CT images.

We propose a method for accelerating chaotic evolution (CE) search using the triple and quadruple comparison mechanisms. We utilize some performance measurements to analyse and verify our proposed algorithm with benchmark functions. The CE is one of evolutionary computation (EC) algorithms that fuses the iteration of evolution and the ergodicity of a chaos system for optimization. We apply the opposition-based learning (OBL) mechanism to CE algorithm to obtain opposite vectors in its search process. Besides the target vectors and chaotic vectors in the conventional CE algorithm, the opposite vectors are also examined during determining the offspring individual for the next generation. Generally, one of drawbacks for the conventional EC algorithm is that premature convergence towards a local optimum instead of a global optimum. The advantage of our proposed algorithm is that it has a higher possibility to avoid being trapped in a premature convergence so that it can reduce a lot of unnecessary computational costs. We also evaluate these algorithms using 12 benchmark functions and some performance measurements. The experiments found that applying OBL mechanism to the CE algorithm can obtain a better optimization performance than the conventional one, especially in the high dimensional optimization tasks.

we propose a regression method using partial least square (PLS) technique and kernel method, which we refer to as kernel partial least square regression II to distinguish the conventional kernel PLS regression. The motivation of the conventional kernel PLS regression is to establish a coordinator system where the independent and dependent variables have a stronger correlation, and which models a regression model using the coordinator system. However, it projects independent variables into a reproducing kernel Hilbert space (RKHS) alone. The proposal extends the basic framework of conventional kernel PLS regression. The proposed method does not only project independent variables into the RKHS, but also dependent variables, and establishes a coordinator system where the independent and dependent variables have a stronger correlation. We use two function regression cases to evaluate the proposed method compared with the conventional kernel PLS regression. The regression performance of the proposed method has almost the same regression accuracy arising from the evaluation result, and this depends on the regression tasks. We explain the correlation calculations of our proposed method, conventional kernel PLS regression, and PLS regression. The meaning of correlation depends on the application in question. We also analyse and discuss the algorithm implementation, correlation meaning, and other issues for further development of the proposal.

We introduce two competitive strategies into conventional differential evolution (DE) to speed up its convergence by increasing competitive pressures among individuals and evaluate the proposals. The first strategy gives individuals with better fitness a higher opportunity for generating more offspring individuals, while conventional DE allows each parent individual to generate only one offspring individual fairly. This strategy compares each of poor individuals with a randomly selected individual from the current population. If the latter becomes a winner, the latter can generate one more offspring individual, but the former loses an opportunity for generating its offspring. If the former becomes a winner, no one loses this opportunity, and each of them generates one offspring individual. The second strategy does not compare a generated offspring individual with its parent but the worst individual in the current population, which can accelerate the elimination of poor individuals and keep better individuals. We design a set of controlled experiments to evaluate these two strategies using CEC2013 benchmark functions with three different dimensions. The experimental results indicate that properly enhancing competition among individuals in DE can speed up its convergence and improve optimization performance.

We propose a method that uses kernel method-based algorithms to implement an autoencoder. Deep learning-based algorithms have two characteristics, one is the high level data abstraction, the other is the multiple level data transformations and representations. The kernel method is one of the approaches that can be used in linear and non-linear transformations. It should be one of the implementations of these transformations in the deep learning. In this paper, the encoder part and decoder part of the autoencoder are implemented by kernel-based principal component analysis and kernel-based linear regression, respectively. As autoencoder is a basic structure and algorithm in deep learning, the proposed method can implement deep learning model and algorithm using duplicate structures. We use image data to evaluate our proposed method. The results show that kernel-based autoencoder can represent and restore image data, but the performance depends on the kernel function and its parameters' selection. We also discuss and analyse some open topics and works towards a study of kernel method-based deep learning.

We use the convergence points estimated by our proposed method as elite individuals for evolutionary computation and evaluate the acceleration effect and analyze the effect and computational cost. The worst individuals in population are replaced with the convergence points estimated from the moving vectors between parent individuals and their offspring; i.e. these convergence points are used as elite individuals. Differential evolution (DE) and 14 benchmark functions are used in our evaluation experiments. The experimental results show that use of the estimated convergence points as elite can accelerate DE search in spite of the calculation cost of the convergence points. We finally analyze the components of the proposed estimation method to improve cost-performance.

We propose a method to solve the selection problem of principal components in machine learning algorithms based on orthogonal transformation by using interactive evolutionary computation. One of the addressed subjects for machine learning algorithms based on orthogonal transformation is how to decide the number of principal components, and which of the principal components should be used to reconstruct the original data. In this work, we use the interactive differential evolution algorithm to study these subjects by using real humans' subjective evaluation in an optimization process. An image compression problem using principal component analysis is introduced to study the proposed method. From the evaluation, we do not only solve the selection problem of principal components for machine learning algorithms based on orthogonal transformation, but also can analyse the human aesthetical characteristics on visual perception and feature selection from the designed method and experimental evaluation. We also discuss and analyse potential research subjects and some open topics, which are invited to further investigate.

We propose methods of estimating the convergence point for the moving vectors of individuals between generations or evolution paths and show that the estimated convergence point can be useful information for accelerating evolutionary computation (EC). As the first stage of this new approach, we do not combine the proposed methods with EC search in this paper, but rather evaluate how power an individual the the estimated convergence point is by comparing fitness values. Through experimental evaluations, we show that the estimated point can be a powerful elite for unimodal fitness landscapes and that clustering moving vectors according to the aimed points is the next research target for multimodal fitness landscape.

We propose a triple comparison-based interactive differential evolution (IDE) algorithm. The comparison of target vector and trail vector supports a local fitness landscape for IDE algorithm to conduct a memetic search. Besides target vector and trail vector in canonical IDE algorithm framework, we conduct a memetic search around whichever is the vector with better fitness. We use a random number from a normal distribution generator or a uniform distribution generator to perturb the vector for generating a third vector. By comparing the target vector, the trail vector and the third vector, we implement a triple comparison mechanism in IDE algorithm. A Gaussian mixture model is applied as a pseudo IDE user in our evaluation. We compare our proposal with canonical IDE and triple comparison-based IDE implemented by opposite-based learning, and apply several statistical tests to investigate the significance of our proposed algorithm. From the evaluation results, our proposed triple comparison-based IDE algorithm shows significantly better performance optimization. We also investigate potential issues arising from our proposal, and discuss some open topics and future opportunities.

We consider algorithmic design, enhancement, and improvement of evolutionary computation (EC) as a mechanism design problem. All individuals or several groups of individuals can be considered as self-interested agents. The individuals in EC can manipulate the parameter settings and operations of an EC algorithm by satisfying their own preferences rather than by following a fixed algorithm rule. EC algorithm designers or EC self-adaptive methods should construct appropriate rules and mechanisms for all agents (individuals) to conduct their evolution behavior correctly in order to definitely achieve the desired and pre-set objective(s) definitively. We propose a formal framework on parameter setting, strategy selection, and algorithmic design of EC by considering the strategy equilibrium implementation of a mechanism design problem in the search process. We attempt to use Nash strategy equilibrium (NE) concept in an implementation of an algorithmic mechanism design problem, but our proposed framework is not limited to Nash strategy equilibrium. The evaluation results present the efficiency of the framework. Its primary principle can be implemented in any EC algorithm that needs to consider the strategy selection issue in its optimization process. The final objective of our work is to implement EC design as an algorithmic mechanism design problem and establish EC fundamental aspects based on this perspective.

We propose a series of data analysis methods for both supervised and un-supervised learning techniques. Three objectives of data relationship and characteristics are used to establish a uniform framework of our proposed methods, which are inspired by principal component analysis and linear discriminant analysis. By using the three objectives and some combinations of them, we investigate and illustrate the performance of the proposed methods. We use simulation data and classical Iris data to investigate the proposed methods. Some discoveries and issues are analysed and discussed arising from the evaluation results. The advantages of the proposed framework do not only depend on its explanation capability of data relationship, but also depend on the fusion of multiple data projection techniques. We investigate some potential research issues of the proposed methods. Some works which extend the current study with kernel method are analysed theoretically. We also present some characteristics of the proposal and discuss some open opportunities and future works.

Ensemble learning system could lessen the degree of overfitting that often appear in the supervised learning process for a single learning model. However, overfitting had still been observed in negative correlation learning that is an ensemble learning method with correlation-based penalty. Two constraints were introduced into negative correlation learning in order to conquer such overfitting. One is the lower bound of error rate (LBER). The other is the upper bound of error output (UBEO). With LBER and UBEO, negative correlation learning will selectively learn the data points. After the performance becomes better than LBER, those unlearned data points with the error output larger than UBEO would not be learned anymore in negative correlation learning. This paper presented the experimental results to explain how these two constraints would affect the performance of negative correlation learning.

Besides the studied transition learning between the two different ensemble learning algorithms such as negative correlation learning and balanced ensemble learning, transition learning could also implemented in negative correlation learning with different correlation penalties. On one hand, negative correlation learning with the lower correlation penalty named as low negative correlation learning might learn too much the training data while generating less negatively correlated neural networks. On the other hand, negative correlation learning with the higher correlation penalty called as high negative correlation learning might not be able to learn the training data, but be capable of generating highly negatively correlated neural networks. By conducting transition learning from low negative correlation learning to high negative correlation learning, this paper shows that the ensembles could have both the good performance and the diverse individual neural networks.

The purpose of our study is to induce compact and high performance machine learning models. In our earlier study, we proposed a decision boundary making (DBM) algorithm. The main philosophy of the DBM algorithm is to reconstruct a high performance model with much smaller cost. In our study, we use support vector machine as a high performance model, and a multilayer neural network, i.e., multilayer perceptron (MLP), as the small model. Experimental results obtained so far show that high performance and compact MLPs can be obtained using DBM. However, there are several parameters of DBM that need to be adjusted appropriately in order to achieve better performance. In this paper, we investigate the effect of parameter N, which is the number of newly generated data, on the performance of obtained MLPs. We discuss the issue that how many new data we should generate to obtain a better performance of DBM. We also investigate the effect of outliers on the performance of the obtained MLPs. Outliers are generally known to be harmful for pattern recognition. Our experimental results show, however, that for some databases, outliers can be useful for obtaining high performance MLPs.

In this paper, we analyse and discuss the relationship between optimization performance of chaotic evolution (CE) algorithm and distribution characteristic of chaotic parameter. CE is an evolutionary computation algorithm that simulates chaotic motion of a chaotic system in a search space for implementing optimization. However, its optimization performance, internal process mechanism and optimization principle are not well studied. In this paper, we investigate distribution characteristics of chaotic systems, which support chaotic parameter in CE algorithm. Compared with other two parameter generators, i.e., a quadratic-like random generator and an uniform random generator, CE algorithm with chaotic parameter generated by the logistic map (μ = 4) shows better optimization performance significantly. We also make an algorithm comparison with differential evolution and an algorithm ranking by Friedman test and Bonferroni-Dunn test. The related topics on relationship between optimization performance of CE algorithm and chaotic parameter distribution are analysed and discussed. From these analyses and discussions, it indicates that chaotic parameter distribution is a significant factor that influences optimization performance of CE algorithm.

We propose a triple comparison and a quadruple comparison-based mechanism for enhancing differential evolution (DE), especially for interactive DE (IDE) where the method can be used to reduce IDE user fatigue. Besides the target vector and trial vector from normal DE, opposition vectors generated by opposition-based learning are used to determine offspring, and the best vector from among these three or four vectors becomes offspring for the next generation. We evaluate the proposed methods by comparing them with conventional IDE and conventional opposition-based IDE using a simulated IDE modeled using a four dimensional Gaussian mixture model. We also evaluate them in DE using 24 benchmark functions. The experiments show that our proposed methods can enhance IDE and DE search efficiently according to several evaluation indices. These include the converged fitness values after the same number of generations, converged fitness values after the same number of fitness calculations, fitness calculation cost, convergence success rates and acceleration rates.

We propose a method for approximating a fitness landscape using adaptive support vector regression (SVR) with opposition based learning (OBL) to enhance the evolutionary search. This method tries to resolve the complexity of the fitness landscape in the original search space by designing a suitable kernel function with an adaptive parameter tuned by OBL, This kernel projects the original search space into a higher dimensional search space with a different topological structure. The elite is obtained from the approximated fitness landscape, using the adaptive SVR to accelerate the evolutionary computation (EC) search, and the individual with the worst fitness is replaced. The merits of the proposed method are evaluated by comparing it with the fitness landscape approximated in the original, in a lower and in a higher dimensional search space.

We propose an approach for approximating a fitness landscape by filtering its frequency components in order to accelerate evolutionary computation (EC) and evaluate the performance of the technique. In addition to the EC individuals, the entire fitness landscape is resampled uniformly. The frequency information for the fitness landscape can then be obtained by applying the discrete Fourier transform (DFT) to the resampled data. Next, we filter to isolate just the major frequency component; thus we obtain a trigonometric function approximating the original fitness landscape after the inverse DFT is applied. The elite is obtained from the approximated function and the EC search accelerated by replacing the worst EC individual with the elite. We use benchmark functions to evaluate some variations of our proposed approach. These variations include the combination of resampling of the global area, local area, in all n-D at once, and in each of n 1-D. The experimental results show that our proposed method is efficient in accelerating most of the benchmark functions.

This paper presents an empirical study on the influence of approximation approaches on accelerating the fireworks algorithm search by elite strategy. In this study, we use three sampling data methods to approximate fitness landscape, i.e. the best fitness sampling method, the sampling distance near the best fitness individual sampling method and the random sampling method. For each approximation methods, we conduct a series of combinative evaluations with the different sampling method and sampling number for accelerating fireworks algorithm. The experimental evaluations on benchmark functions show that this elite strategy can enhance the fireworks algorithm search capability effectively. We also analyze and discuss the related issues on the influence of approximation model, sampling method, and sampling number on the fireworks algorithm acceleration performance.