Optimal Values Of Hyperparameters Research Articles

PM2.5 Concentration Estimation Using Bi-LSTM with Osprey Optimization Method

Outdoor air pollution causes a lot of health problems for humans. Particulate Matter 2.5 (PM2.5), due to its small size, can enter the human respiratory system with ease and cause significant health effects on humans. This makes PM2.5 significant among the various air pollutants. Hence, it is important to measure the value of PM2.5 accurately for better management of air quality. Algorithms for deep learning and machine learning can be used to forecast air quality data. A model that minimizes the prediction error of the PM2.5 forecast is needed. In this paper, a PM2.5 concentration estimation model using Bi-LSTM (Bidirectional Long Short-Term Memory) with meteorological data as predictor variables is proposed. For a better estimation of PM2.5 values, the hyperparameters of the Bi-LSTM model used are tuned using the Osprey Optimization Algorithm (OOA), a recent meta-heuristic algorithm. The model that works with the optimal values of hyperparameters identified by OOA performed better than the other models when they are compared based on evaluation metrics like Mean-Squared Error and R2.

An Advanced Methodology for Crystal System Detection in Li-ion Batteries

Detecting the crystal system of lithium-ion batteries is crucial for optimizing their performance and safety. Understanding the arrangement of atoms or ions within the battery’s electrodes and electrolyte allows for improvements in energy density, cycling stability, and safety features. This knowledge also guides material design and fabrication techniques, driving advancements in battery technology for various applications. In this paper, a publicly available dataset was utilized to develop mathematical equations (MEs) using a genetic programming symbolic classifier (GPSC) to determine the type of crystal structure in Li-ion batteries with a high classification performance. The dataset consists of three different classes transformed into three binary classification datasets using a one-versus-rest approach. Since the target variable of each dataset variation is imbalanced, several oversampling techniques were employed to achieve balanced dataset variations. The GPSC was trained on these balanced dataset variations using a five-fold cross-validation (5FCV) process, and the optimal GPSC hyperparameter values were searched for using a random hyperparameter value search (RHVS) method. The goal was to find the optimal combination of GPSC hyperparameter values to achieve the highest classification performance. After obtaining MEs using the GPSC with the highest classification performance, they were combined and tested on initial binary classification dataset variations. Based on the conducted investigation, the ensemble of MEs could detect the crystal system of Li-ion batteries with a high classification accuracy (1.0).

Additive-based concrete evaluation using categorical boosting method

This study utilized several coupled approaches to create powerful algorithms for forecasting the compressive strength (Cs) of concretes that include metakaolin (MK) and fly ash (FA). For this purpose, three various methods were considered, named random forests (RF), Categorical boosting model (CB), and extreme gradient boosting (XGB) by considering the seven most influential input variables. It was tried to divide the concrete components to binder value (B) to achieve the non-dimensional input variables. Herein, the cutting-edge Tasmanian devil Optimization (TDO) algorithm was linked with RF, XGB, and CB for the purpose of determining the optimal values of hyperparameters (named TD - CB, TD - RF, and TD - XG). It is worth mentioning that developing the mentioned algorithms optimized with TD to estimate the mechanical properties of the concrete containing several important admixtures can be recognized as this study’s contribution to practical applications. The findings indicate that the algorithms possess a notable capacity to precisely forecast the Cs of concrete, which includes MK and FA, with R2 bigger than roughly 0.97. The lower value of OBJ comprehensive index belonged to the TD - CB at 1.5762, followed by TD - XG at 1.9943 and then 2.3317 related to TD - RF with almost 70% reduction. The sensitivity analysis demonstrated that the prediction of Cs is highly influenced by all input parameters, which are higher than 0.8659, but a higher influence from MK/B at 0.9548.

Two-Stage Neural Network Optimization for Robust Solar Photovoltaic Forecasting

Neural networks (NNs) have shown outstanding performance in solar photovoltaic (PV) power forecasting due to their ability to effectively learn unstable environmental variables and their complex interactions. However, NNs are limited in their practical industrial application in the energy sector because the optimization of the model structure or hyperparameters is a complex and time-consuming task. This paper proposes a two-stage NN optimization method for robust solar PV power forecasting. First, the solar PV power dataset is divided into training and test sets. In the training set, several NN models with different numbers of hidden layers are constructed, and Optuna is applied to select the optimal hyperparameter values for each model. Next, the optimized NN models for each layer are used to generate estimation and prediction values with fivefold cross-validation on the training and test sets, respectively. Finally, a random forest is used to learn the estimation values, and the prediction values from the test set are used as input to predict the final solar PV power. As a result of experiments in the Incheon area, the proposed method is not only easy to model but also outperforms several forecasting models. As a case in point, with the New-Incheon Sonae dataset—one of three from various Incheon locations—the proposed method achieved an average mean absolute error (MAE) of 149.53 kW and root mean squared error (RMSE) of 202.00 kW. These figures significantly outperform the benchmarks of attention mechanism-based deep learning models, with average scores of 169.87 kW for MAE and 232.55 kW for RMSE, signaling an advance that is expected to make a significant contribution to South Korea’s energy industry.

Detection and Classification of Brain Tumor Using Convolution Extreme Gradient Boosting Model and an Enhanced Salp Swarm Optimization

Some types of tumors in people with brain cancer grow so rapidly that their average size doubles in twenty-five days. Precisely determining the type of tumor enables physicians to conduct clinical planning and estimate dosage. However, accurate classification remains a challenging task due to the variable shape, size, and location of the tumors.The major objective of this paper is to detect and classify brain tumors. This paper introduces an effective Convolution Extreme Gradient Boosting model based on enhanced Salp Swarm Optimization (CEXGB-ESSO) for detecting brain tumors, and their types. Initially, the MRI image is fed to bilateral filtering for the purpose of noise removal. Then, the de-noised image is fed to the CEXGB model, where Extreme Gradient Boosting (EXGB) is used, replacing a fully connected layer of CNN to detect and classify brain tumors. It consists of numerous stacked convolutional neural networks (CNN) for efficient automatic learning of features, which avoids overfitting and time-consuming processes. Then, the tumor type is predicted using the EXGB in the last layer, where there is no need to bring the weight values from the fully connected layer. Enhanced Salp Swarm Optimization (ESSO) is utilized to find the optimal hyperparameters of EXGB, which enhance convergence speed and accuracy. Our proposed CEXGB-ESSO model gives high performance in terms of accuracy (99), sensitivity (97.52), precision (98.2), and specificity (97.7).Also, the convergence analysis reveals the efficient optimization process of ESSO, obtaining optimal hyperparameter values around iteration 25. Furthermore, the classification results showcase the CEXGB-ESSO model’s capability to accurately detect and classify brain tumors.

A novel approach for prediction of groundwater quality using gradient boosting-based algorithms

ABSTRACT This study explores the potential of the GPBoost approach for groundwater quality assessment in comparison to three other gradient boosting-based algorithms. Three methods, random search, grid search, and Bayesian optimization were used to find the optimal values of various hyperparameters with all four-gradient boosting-based algorithms. One hundred and two samples of Entropy weighted water quality index with 14 input parameters are used for assessing groundwater quality. The calculated EWQI values for drinking range between 80.4 and 394.96 in pre-monsoon and 39.6 to 338.79 during the post-monsoon period. Moreover, spatial distribution maps displayed that the central portions of the study area fall under medium water quality. The performances of models were compared based on multiple statistical criteria, including Correlation Coefficient (CC), root mean square error (RMSE), and mean absolute error (MAE). The results reveal that the CC value by all modeling approaches is more than 0.93, suggesting a comparable performance by all methods. Results in terms of RMSE values in predicting the EWQI values suggest GPBoost (random search) model performed better than the other three models, thus suggesting a competitive performance by GPBoost in comparison to other gradient boosting-based approaches. Relative importance analysis provided by random and grid search methods highlights the significance of NO3 −, Mg2+, TDS, EC, and TH as important input parameters for predicting EWQI.

Robust sign language detection for hearing disabled persons by Improved Coyote Optimization Algorithm with deep learning

<abstract> <p>Sign language (SL) recognition for individuals with hearing disabilities involves leveraging machine learning (ML) and computer vision (CV) approaches for interpreting and understanding SL gestures. By employing cameras and deep learning (DL) approaches, namely convolutional neural networks (CNN) and recurrent neural networks (RNN), these models analyze facial expressions, hand movements, and body gestures connected with SL. The major challenges in SL recognition comprise the diversity of signs, differences in signing styles, and the need to recognize the context in which signs are utilized. Therefore, this manuscript develops an SL detection by Improved Coyote Optimization Algorithm with DL (SLR-ICOADL) technique for hearing disabled persons. The goal of the SLR-ICOADL technique is to accomplish an accurate detection model that enables communication for persons using SL as a primary case of expression. At the initial stage, the SLR-ICOADL technique applies a bilateral filtering (BF) approach for noise elimination. Following this, the SLR-ICOADL technique uses the Inception-ResNetv2 for feature extraction. Meanwhile, the ICOA is utilized to select the optimal hyperparameter values of the DL model. At last, the extreme learning machine (ELM) classification model can be utilized for the recognition of various kinds of signs. To exhibit the better performance of the SLR-ICOADL approach, a detailed set of experiments are performed. The experimental outcome emphasizes that the SLR-ICOADL technique gains promising performance in the SL detection process.</p> </abstract>

Моделювання автономного проходження лабіринту з використанням алгоритму NEAT

The article deals with the problem of finding a solution for the navigational task of navigating a maze by an autonomous agent controlled by an artificial neural network (ANN). A solution to this problem was proposed by training the controlling ANN using the method of neuroevolution of augmenting topologies (NEAT). A description of the mathematical apparatus for determining the goal-oriented objective function to measure fitness of the decision-making agent, suitable for optimizing the training of ANN in the process of neuroevolution, was given. Based on the invented objective function, a software was developed to control the neuroevolutionary process using the Python programming language. A system for simulating the behavior of an autonomous robot that can navigate through a maze using input signals from various types of sensors has been created. The simulation system allows to imitate the behavior of a physical robot in a large number of experiments in a short time and with minimal expenses. The experiments performed using the created simulation system to find the optimal values of hyperparameters, which can be used for successful training of the controlling ANN by the method of neuroevolution, are presented. Additionally, the implemented new methods of visualizing the training process are described. These methods significantly simplify the search for optimal hyperparameters of the NEAT algorithm, due to the visual demonstration of the effect of changing one or another parameter on the training process.

Development of hybrid models based on deep learning and optimized machine learning algorithms for brain tumor Multi-Classification

Accurate classification of magnetic resonance imaging (MRI) images of brain tumors is crucial for early diagnosis and effective treatment in clinical studies. In these studies, many models supported by artificial intelligence (AI) have been proposed as assistant systems for experts. In particular, state-of-the-art deep learning (DL) models that have proven themselves in different fields have been effectively used in the classification of brain MRI images. However, the low accuracy of multiple classification of these images still leads researchers to conduct different studies in this field. Especially there is a need to develop models that achieve high accuracy on original images, and it is believed that this need can be met not only by DL models but also by classical machine learning (ML) algorithms. However, it is critical to choose the hyperparameters correctly for the hybrid use of ML algorithms with DL models. This study proposes a powerful new hybrid method to perform multiple classifications of brain tumors with high accuracy. This method also uses a novel convolutional neural network (CNN) model for feature extraction, and ML algorithms are used for feature classification. In addition, nine state-of-the-art CNN models are used for CNN performance comparison. The Bayesian optimization algorithm is used to obtain the optimal hyperparameter values of ML algorithms. The results obtained from the experimental studies show that the proposed hybrid model achieved 97.15% mean classification accuracy and 97% recall, precision, and F1-score values. Other hybrid models, including DarkNet19-SVM, DarkNet53-SVM, DenseNet201-SVM, EfficientNetB0-SVM, InceptionV3-SVM, NasNetMobile-SVM, ResNet50-SVM, ResNet101-SVM, and Xception-SVM, achieved mean classification accuracies of 95.01%, 95.58%, 96.87%, 97.01%, 95.3%, 95.01%, 96.3%, 95.87%, and 96.23%, respectively. Additionally, the proposed hybrid model exhibited remarkable time efficiency, accomplishing the classification process in a mere 67 min. Conversely, the model that exhibited the lowest time efficiency was the InceptionV3, with a processing time of 370 min. In terms of computational complexity, the EfficientNetB0 model is the most efficient. Despite the higher computational complexity of the proposed CNN model compared to some other models, it achieves the second-best classification accuracy. These results show that the proposed method performs better than previous studies on the same dataset. Especially in the classification problem, the optimized ML algorithms were superior to CNN classifiers. Finally, except for one, the proposed CNN model achieved better classification accuracies than the state-of-the-art CNN models.

Transfer Learning for P300 Brain-Computer Interfaces by Joint Alignment of Feature Vectors.

This article presents a new transfer learning method named group learning, that jointly aligns multiple domains (many-to-many) and an extension named fast alignment that aligns any further domain to previously aligned group of domains (many-to-one). The proposed group alignment algorithm (GALIA) is evaluated on brain-computer interface (BCI) data and optimal hyper-parameter values of the algorithm are studied for classification performance and computational cost. Six publicly available P300 databases comprising 333 sessions from 177 subjects are used. As compared to the conventional subject-specific train/test pipeline, both group learning and fast alignment significantly improve the classification accuracy except for the database with clinical subjects (average improvement: 2.12±1.88%). GALIA utilizes cyclic approximate joint diagonalization (AJD) to find a set of linear transformations, one for each domain, jointly aligning the feature vectors of all domains. Group learning achieves a many-to-many transfer learning without compromising the classification performance on non-clinical BCI data. Fast alignment further extends the group learning for any unseen domains, allowing a many-to-one transfer learning with the same properties. The former method creates a single machine learning model using data from previous subjects and/or sessions, whereas the latter exploits the trained model for an unseen domain requiring no further training of the classifier.

An optimized deep learning model for human activity recognition using inertial measurement units

AbstractHuman activity recognition (HAR) has recently gained popularity due to its applications in healthcare, surveillance, human‐robot interaction, and various other fields. Deep learning (DL)‐based models have been successfully applied to the raw data captured through inertial measurement unit (IMU) sensors to recognize multiple human activities. Despite the success of DL‐based models in human activity recognition, feature extraction remains challenging due to class imbalance and noisy data. Additionally, selecting optimal hyperparameter values for DL models is essential since they affect model performance. The hyperparameter values of some of the existing DL‐based HAR models are chosen randomly or through the trial‐and‐error method. The random selection of these significant hyperparameters may be suitable for some applications, but sometimes it may worsen the model's performance in others. Hence, to address the above‐mentioned issues, this research aims to develop an optimized DL model capable of recognizing various human activities captured through IMU sensors. The proposed DL‐based HAR model combines convolutional neural network (CNN) layers and bidirectional long short‐term memory (Bi‐LSTM) units to simultaneously extract spatial and temporal sequence features from raw sensor data. The Rao‐3 metaheuristic optimization algorithm has been adopted to identify the ideal hyperparameter values for the proposed DL model in order to enhance its recognition performance. The proposed DL model's performance is validated on PAMAP2, UCI‐HAR, and MHEALTH datasets and achieved 94.91%, 97.16%, and 99.25% accuracies, respectively. The results reveal that the proposed DL model performs better than the existing state‐of‐the‐art (SoTA) models.

Breast Cancer Classification Using Synthesized Deep Learning Model with Metaheuristic Optimization Algorithm.

Breast cancer is the second leading cause of mortality among women. Early and accurate detection plays a crucial role in lowering its mortality rate. Timely detection and classification of breast cancer enable the most effective treatment. Convolutional neural networks (CNNs) have significantly improved the accuracy of tumor detection and classification in medical imaging compared to traditional methods. This study proposes a comprehensive classification technique for identifying breast cancer, utilizing a synthesized CNN, an enhanced optimization algorithm, and transfer learning. The primary goal is to assist radiologists in rapidly identifying anomalies. To overcome inherent limitations, we modified the Ant Colony Optimization (ACO) technique with opposition-based learning (OBL). The Enhanced Ant Colony Optimization (EACO) methodology was then employed to determine the optimal hyperparameter values for the CNN architecture. Our proposed framework combines the Residual Network-101 (ResNet101) CNN architecture with the EACO algorithm, resulting in a new model dubbed EACO-ResNet101. Experimental analysis was conducted on the MIAS and DDSM (CBIS-DDSM) mammographic datasets. Compared to conventional methods, our proposed model achieved an impressive accuracy of 98.63%, sensitivity of 98.76%, and specificity of 98.89% on the CBIS-DDSM dataset. On the MIAS dataset, the proposed model achieved a classification accuracy of 99.15%, a sensitivity of 97.86%, and a specificity of 98.88%. These results demonstrate the superiority of the proposed EACO-ResNet101 over current methodologies.

Performance evaluation of deep learning algorithms for heat loss damage classification in buildings from UAV-borne infrared images

Deep learning (DL) algorithms are gaining popularity to automate anomaly detection in a variety of civil and mechanical engineering domains. Nevertheless, very few attempts have been made to use infrared (IR) images with DL algorithms to detect heat loss in a building and expedite the otherwise cumbersome energy audit process. However, no clear guidelines exist for the selection of appropriate algorithms or hyperparameters (e.g., number of epochs, learning rate, number of images) to implement DL with IR images. To address the research gap, a comparative sensitivity analysis of four well-known DL algorithms for the classification of heat losses using only IR images was performed for the first time. Four DL algorithms, namely, general convolutional neural network (CNN), VGG16, transfer learning (TL) VGG16, and TL inceptionV3 were selected and used for automating the detection of heat loss from IR images. The sensitivity of the selected CNN algorithms was analyzed based on i) the size of the IR images dataset, ii) the number of epochs, and iii) the effect of an unbalanced dataset. The goal was to provide a framework that may aid in the selection of suitable CNN architecture as well as parameters related to data size and an optimal number of training iterations for reliable heat loss identification using IR images. This article has provided a framework for automated heat loss detection that can be applied to any IR dataset without the need to pre-process the IR images. The accuracy of the selected DL algorithms is validated with IR images of laboratory setups and real-world buildings characterized by different types of defects. The comparative sensitivity analysis was conducted based on multiple performance indicators such as precision, recall, and F-1 score which resulted in optimal values of hyperparameters and the most suitable DL algorithm. The automated identification of heat loss yielded performance indicator values of more than 90%.

Hybrid Machine Learning Approach for Evapotranspiration Estimation of Fruit Tree in Agricultural Cyber-Physical Systems.

The flourish of the Internet of Things (IoT) and data-driven techniques provide new ideas for enhancing agricultural production, where evapotranspiration estimation is a crucial issue in crop irrigation systems. However, tremendous and unsynchronized data from agricultural cyber-physical systems bring large computational costs as well as complicate performing conventional machine learning methods. To precisely estimate evapotranspiration with acceptable computational costs under the background of IoT, we combine time granulation computing techniques and gradient boosting decision tree (GBDT) with Bayesian optimization (BO) to propose a hybrid machine learning approach. In the combination, a fuzzy granulation method and a time calibration technique are introduced to break voluminous and unsynchronized data into small-scale and synchronized granules with high representativeness. Subsequently, GBDT is implemented to predict evapotranspiration, and BO is utilized to find the optimal hyperparameter values from the reduced granules. IoT data from Xi'an Fruit Technology Promotion Center in Shaanxi Province, China, verify that the proposed granular-GBDT-BO is effective for cherry tree evapotranspiration estimation with reduced computational time, and acceptable and robust predictive accuracy. Consequently, the precise estimation of crop evapotranspiration could provide operational guidance for plant irrigation, plant conservations, and pest control in the agricultural greenhouse.

Enhancing a convolutional neural network model for land subsidence susceptibility mapping using hybrid meta-heuristic algorithms

Managing natural hazards such as land subsidence (LS) is important because they cause large economic and human loss. LS has become a significant challenge in South Korea due to its many abandoned coal mines. Therefore, preparing LS zoning maps is vital to controlling damage caused by LS. In this study, genetic algorithm (GA) and binary whale optimization algorithm were used to select the most influential conditioning factors from the initial set of 19 factors. Subsequently, the whale optimization algorithm, and Laplacian whale optimization algorithm (LXWOA) were utilized to determine the optimal hyperparameter values for the convolutional neural network (CNN) model. The results showed that the CNN-GA-LXWOA algorithm provided a more accurate and reliable LS susceptibility map, with an area under the receiver operating characteristic curve of 0.9606, root mean square error of 0.2974 and accuracy of 91.09%. This algorithm covered 95.6% of past subsidence occurrences in the high and very high LS susceptibility classes, demonstrating its suitability for predicting future subsidence areas. We found that six main factors (elevation, drift, lineament density, slope, distance to railroads, and railroad density) control LS occurrence in Taebaek, followed by groundwater depth, lithology, and profile curvature. The proposed model may be applied for other regions with different parameters and environmental factors due to its flexible structure. The final map created in this study provides a useful tool for better LS management to mitigate the adverse effects of this natural hazard in the study area.

Breast cancer diagnosis using Orca predation optimization algorithm

Early detection and classification of breast cancer can be facilitated to initiate the most effective treatment. As the second leading cause of death among women, early breast cancer screening is essential for reducing mortality rates. In this context, Convolutional neural networks (CNNs) are the ideal candidate for increasing the rate of identification and classification of tumours with efficiency, particularly in medical imaging. This research proposes a hybridised CNN with the Orca Predation Optimization Algorithm (OPOA) as a novel classification model for the effective detection of abnormalities in breast cancer diagnosis. Specifically, the OPOA technique is used to determine the optimal hyperparameter values for the hybrid CNN architecture being deployed. As the pretrained CNN model, the suggested model utilizeds a ResNet50 residual network. It merged OPOA with the ResNet50 residual network to construct the OPOA-ResNet-50 Architecture. The experimental validation of the proposed OPOA-ResNet-50 model utilising the datasets of curated breast imaging subset of DDSM (CBIS-DDSM) shown improved classification accuracy of 99.04%, specificity of 98.56%, and sensitivity of 97.78% in comparison to the baseline techniques. The results also revealed that the proposed under mammographic image analysis society (MIAS) OPOA-ResNet-50 model demonstrated superior classification accuracy of 98.64%, specificity of 98.79%, and sensitivity of 98.82% compared to the benchmarked methods. The adopted OPOA algorithm is determined to achieve more optimal hyperparameter values for the ResNet50 architecture than the comparative algorithms Improved Marine Predator Optimization Algorithm (IMPOA), Whale Optimization Algorithm (WOA), Harris hawk’s optimization (HHO), and gravitational search algorithm (GSA).

Applications of fractal interpolants in kernel regression estimations

Generating a function from its finite set of samples is widely studied in data fitting problems. The theory of nonparametric modeling has been developed by many researchers, and several types of estimators have been established in the literature. On the other hand, the fractal theory provides new technologies for making complicated curves and fitting experimental data, and fractal interpolation functions are established as generalizations of traditional interpolation techniques. The aim of this paper is to investigate applications of (smooth) fractal interpolation functions in nonparametric kernel regressions. We discuss a key sufficient condition in detail in the construction of (smooth) fractal interpolation functions. Then the Priestley–Chao estimator is considered, and we establish its fractal perturbation. The raw data we consider in our example is the Crude Oil WTI Futures daily highest price from 2021/10/13 to 2022/7/29. We take 11 samples from the raw data to construct a fractal perturbation of the Priestley–Chao regression function. SLSQP is applied to find the optimal values of hyperparameters so that the given empirical mean squared error is minimum. We see that nonparametric kernel regressions for finite sample data are widely used in applications, and we can decrease the empirical mean squared error further by considering fractal perturbations of these regression functions.

HYPERPARAMETER TUNING ON RANDOM FOREST FOR DIAGNOSE COVID-19

Diagnosis of Covid using the RT-PCR (Reverse Transcription Polymerase Chain Reaction) test requires high costs and takes a long time. For this reason, another method is needed that can be used to diagnose Covid-19 quickly and accurately. Random Forest is one of the popular classification algorithms for making predictive models. Random forest involves many hyperparameters that control the structure of each tree, the forest, and its randomness. Random Forest is a method which very sensitive to hyperparameter values, as their prediction accuracy can increase significantly when optimized hyperparameters are predefined and then adjusted according to the procedure. The purpose of doing hyperparameter tuning on the random forest algorithm is to increase accuracy in the diagnosis of covid-19. Searching for optimal values of hyperparameters is done by the Grid Search method and Random Search. The result explains that the Random Forest can be used to diagnose Covid-19 with an accuracy of 94%, and with hyperparameter tuning, it can increase the accuracy of the random forest by 2%.

Modeling and optimization of strategies for making individual decisions in multi-agent socio-economic systems with the use of machine learning

This article presents a new approach to modeling and optimizing individual decision-making strategies in multi-agent socio-economic systems (MSES). This approach is based on the synthesis of agent-based modeling methods, machine learning and genetic optimization algorithms. A procedure for the synthesis and training of artificial neural networks (ANNs) that simulate the functionality of MSES and provide an approximation of the values of its objective characteristics has been developed. The feature of the two-step procedure is the combined use of particle swarm optimization methods (to determine the optimal values of hyperparameters) and the Adam machine learning algorithm (to compute weight coefficients of the ANN). The use of such ANN-based surrogate models in parallel multi-agent real-coded genetic algorithms (MA-RCGA) makes it possible to raise substantially the time-efficiency of the evolutionary search for optimal solutions. We have conducted numerical experiments that confirm a significant improvement in the performance of MA-RCGA, which periodically uses the ANN-based surrogate-model to approximate the values of the objective and fitness functions. A software framework has been designed that consists of the original (reference) agent-based model of trade interactions, the ANN-based surrogate model and the MA-RCGA genetic algorithm. At the same time, the software libraries FLAME GPU, OpenNN (Open Neural Networks Library), etc., agent-based modeling and machine learning methods are used. The system we developed can be used by responsible managers.

Exploring Dimensionality Reduction Techniques for Deep Learning Driven QSAR Models of Mutagenicity.

Dimensionality reduction techniques are crucial for enabling deep learning driven quantitative structure-activity relationship (QSAR) models to navigate higher dimensional toxicological spaces, however the use of specific techniques is often arbitrary and poorly explored. Six dimensionality techniques (both linear and non-linear) were hence applied to a higher dimensionality mutagenicity dataset and compared in their ability to power a simple deep learning driven QSAR model, following grid searches for optimal hyperparameter values. It was found that comparatively simpler linear techniques, such as principal component analysis (PCA), were sufficient for enabling optimal QSAR model performances, which indicated that the original dataset was at least approximately linearly separable (in accordance with Cover's theorem). However certain non-linear techniques such as kernel PCA and autoencoders performed at closely comparable levels, while (especially in the case of autoencoders) being more widely applicable to potentially non-linearly separable datasets. Analysis of the chemical space, in terms of XLogP and molecular weight, uncovered that the vast majority of testing data occurred within the defined applicability domain, as well as that certain regions were measurably more problematic and antagonised performances. It was however indicated that certain dimensionality reduction techniques were able to facilitate uniquely beneficial navigations of the chemical space.