Number Of Epochs Research Articles

Tree Species Classification Based on PointNet++ and Airborne Laser Survey Point Cloud Data Enhancement

Compared with ground-based light detection and ranging (LiDAR) data, the differential distribution of the quantity and quality of point cloud data from airborne LiDAR poses difficulties for tree species classification. To verify the feasibility of using the PointNet++ algorithm for point cloud tree species classification with airborne LiDAR data, we selected 11 tree species from the Minjiang River Estuary Wetland Park in Fuzhou City and Sanjiangkou Ecological Park. Training and testing sets were constructed through pre-processing and segmentation, and direct and enhanced down-sampling methods were used for tree species classification. Experiments were conducted to adjust the hyperparameters of the proposed algorithm. The optimal hyperparameter settings used the multi-scale sampling and grouping (MSG) method, down-sampling of the point cloud to 2048 points after enhancement, and a batch size of 16, which resulted in 91.82% classification accuracy. PointNet++ could be used for tree species classification using airborne LiDAR data with an insignificant impact on point cloud quality. Considering the differential distribution of the point cloud quantity, enhanced down-sampling yields improved the classification results compared to direct down-sampling. The MSG classification method outperformed the simplified sampling and grouping classification method, and the number of epochs and batch size did not impact the results.

Reliability of the P3 component in a 3-stimulus concealed information test with known and unknown stimuli

The parietal P3 amplitude of the event-related potential (ERP) has been highlighted to signal stimulus salience in concealed information tests (CITs) with known and unknown stimuli. To extend previous validity investigations in deception research, the reliability of the early parietal P3 amplitude in a 3-stimulus CIT including known and unknown stimuli has been investigated in a sample of N = 68 participants. Variations of Cronbach’s Alpha and split-half reliability (odd–even, 1st vs. 2nd half) of the early parietal P3 amplitude were considered for ERP quantification method (peak-to-peak, mean, baseline-to-peak), epoch selection strategies (chronological vs. random) and number of epochs (10–40) in three stimulus types (probe, target, irrelevant). Excellent reliability coefficients have been summarized for number of epochs, P3 quantifications, and epoch selection strategies. Moreover, the best-practice recommendations highlight mandatory combinations of epoch selection strategy, number of epochs and P3 quantifications to ensure robust and excellent reliabilities of the early parietal P3 amplitude in a 3-stimulus CIT with known (probe, target) and unknown (irrelevant) stimuli. The robustness of mandatory excellent reliability coefficients for the early parietal P3 has been discussed for two data tracks in single-case analyses.

Bayesian optimization of multi-layer perceptron models for power distribution system state estimation

Abstract Feedforward multilayer perceptron models (MLPs) have been applied to power distribution system state estimation (DSSE) in the past. Existing methods usually employ an ad-hoc or trial and error approach to MLP hyperparameter selection, and thus a systematic way of selecting the optimal hyperparameters including the number of neurons per hidden layer, learning rate, number of training epochs and training batch size is desirable and needed. This paper presents an approach based on Bayesian Optimization with Gaussian Processes for selecting MLP model hyperparameters for state estimation purposes. Results of the optimized MLP models are presented alongside the unoptimized models to compare performance of training, testing, and validation in terms of root-mean-squared-error (RMSE). Additionally, machine learning pipelines were employed and total execution time (seconds) for each trial is presented. The study shows that the MLP models obtained through the proposed optimization method outperform unoptimized models in terms of generalization capability for unseen, new cases.

A novel experimental and machine learning model to remove COD in a batch reactor equipped with microalgae

By using microorganisms and the microalgae Chlorella vulgaris in conjunction with sequencing batch reactors (SBRs), the performance of a wastewater treatment facility was studied. For this purpose, the effect of pH, temperature, {mathrm{COD}}_{mathrm{inlet}}, and air flowrate on COD removal rate and residual was investigated. A single-factorial optimization method is utilized to optimize the amount of COD removal, and the best result is obtained with a pH of 8, {mathrm{COD}}_{mathrm{inlet}}=600, mathrm{mg}/mathrm{l}, and an airflow rate of 55 l/min. Under optimal conditions, the amount of residual COD in the effluent reached 36 mathrm{mg}/mathrm{l}, showing an augmentation in the efficiency of the desired system. Moreover, empirical correlations are proposed for double-factorial optimization of residual COD and COD removal. Also, a multilayer perceptron artificial neural network is proposed to model the process and predict the residual COD concentration. The useful technique of hyperparameter tuning is utilized to obtain the best result for the predictions. All the effective parameters, including the number of hidden layers, neurons, epochs, and batch size, are adjusted. Data from the experiments agreed well with the artificial neural network modeling results. For this modeling, the values of the correlation coefficient ({R}^{2}) and mean absolute error (MAE) were obtained as 0.98 and 2%, respectively.

Pedestrian Detection System using YOLOv5 for Advanced Driver Assistance System (ADAS)

The technology in transportation is continuously developing due to reaching the self-driving vehicle. The need of detecting the situation around vehicles is a must to prevent accidents. It is not only limited to the conventional vehicle in which accident commonly happens, but also to the autonomous vehicle. In this paper, we proposed a detection system for recognizing pedestrians using a camera and minicomputer. The approach of pedestrian detection is applied using object detection method (YOLOv5) which is based on the Convolutional Neural Network. The model that we proposed in this paper is trained using numerous epochs to find the optimum training configuration for detecting pedestrians. The lowest value of object and bounding box loss is found when it is trained using 2000 epochs, but it needs at least 3 hours to build the model. Meanwhile, the optimum model’s configuration is trained using 1000 epochs which has the biggest object (1.49 points) and moderate bounding box (1.5 points) loss reduction compared to the other number of epochs. This proposed system is implemented using Raspberry Pi4 and a monocular camera and it is only able to detect objects for 0.9 frames for each second. As further development, an advanced computing device is needed due to reach real-time pedestrian detection. 

An Invasive Ductal Carcinomas Breast Cancer Grade Classification Using an Ensemble of Convolutional Neural Networks.

Invasive Ductal Carcinoma Breast Cancer (IDC-BC) is the most common type of cancer and its asymptomatic nature has led to an increased mortality rate globally. Advancements in artificial intelligence and machine learning have revolutionized the medical field with the development of AI-enabled computer-aided diagnosis (CAD) systems, which help in determining diseases at an early stage. CAD systems assist pathologists in their decision-making process to produce more reliable outcomes in order to treat patients well. In this work, the potential of pre-trained convolutional neural networks (CNNs) (i.e., EfficientNetV2L, ResNet152V2, DenseNet201), singly or as an ensemble, was thoroughly explored. The performances of these models were evaluated for IDC-BC grade classification using the DataBiox dataset. Data augmentation was used to avoid the issues of data scarcity and data imbalances. The performance of the best model was compared to three different balanced datasets of Databiox (i.e., 1200, 1400, and 1600 images) to determine the implications of this data augmentation. Furthermore, the effects of the number of epochs were analysed to ensure the coherency of the most optimal model. The experimental results analysis revealed that the proposed ensemble model outperformed the existing state-of-the-art techniques in relation to classifying the IDC-BC grades of the Databiox dataset. The proposed ensemble model of the CNNs achieved a 94% classification accuracy and attained a significant area under the ROC curves for grades 1, 2, and 3, i.e., 96%, 94%, and 96%, respectively.

A Meta-Learning Based Framework for Cell-Level Mobile Network Traffic Prediction

In this paper, we propose a meta-learning based cell-level network traffic prediction framework (ML-TP), which can provide the proper initial weight vector for the learning model of a new prediction task based on the prediction task’s meta-features. In the ML-TP, each prediction task forms a base-task, and a multi-layer long short-term memory (LSTM) network is constructed as its base-learner. Through fast Fourier transform (FFT) analyses of real-world network traffic data, we find that the five most dominating frequency components can capture the cell-level traffic variations, and hence can be used as a base-task’s meta-features. We prove that the well-trained weight vector of a previous base-task’s base-learner is likely to be a proper initial weight vector of a new base-task’s base-learner if the meta-features of the two base-tasks are close to each other in the Euclidean space. Accordingly, we propose a K-nearest neighbours (KNN) algorithm based meta-learner to deal with the meta-task in the ML-TP. Numerical tests show that the ML-TP can significantly increase the base-learners’ after-training prediction accuracy and learning efficiency in terms of the number of base-samples and the number of epochs needed in each base-learner’s fine-tuning progress.

On classification of Sentinel-2 satellite images by a neural network ResNet-50

Various combinations of neural network parameters and sets of input data for satellite image classification are considered in the article. The training set is completed with a NDVI (normalized difference vegetation index) and local binary patterns. Testing of classifiers created on a different number of epochs and samples is carried out. Values of the neural network hyperparameters are determined that allow a classification accuracy of 0.70 and an F-measure of 0.65 to be achieved. Separation into classes with similar spectral characteristics is shown to offer low classification quality at different parameters and input data sets. Additional information is required. For example, for forests to be divided into more detailed classes, one needs to employ classifiers that use images from different seasons and vegetation periods. In addition, the training set needs to be extended to take into account various natural zones, soils, etc.

Skin Lesion Classification Using CNN-based Transfer Learning Model

The computer-aided diagnosis (CAD) and the analysis of skin lesions using deep learning models have become common in the last decade. The proposed CAD systems have considered various datasets and deep learning models. The transfer of knowledge from particular pre-trained models to others has also gained importance due to the efficient convergence and superior results. This study presents the design and implementation of a transfer learning model using Convolutional Neural Networks (CNN) with variable training epoch numbers to classify skin lesion images obtained by smartphones. The model is divided into the inner and external CNN models to train and transfer the knowledge, and the preprocessing and data augmentation are not applied. Several experiments are performed to classify cancerous and non-cancerous skin lesions and all skin lesion types provided in the dataset separately. The designed model increased the classification rates by 20% compared to the conventional CNN. The transfer learning model achieved 0.81, 0.88, and 0.86 mean recall, mean specificity, and mean accuracy in detecting cancerous lesions, and 0.83, 0.90, and 0.86 macro recall, macro precision, and macro F1 score in classifying six skin lesions. The obtained results show the efficacy of transfer learning in skin lesion diagnosis.

Stock price prediction with long short-term memory

Stock forecasting aims to predict future stock prices based on past price changes in the market, playing an essential role in the field of financial transactions. However, since the stock market is highly uncertain, stock prediction is complex and challenging. This paper uses the long short-term memory (LSTM) model to predict the stock market and compares it with the current stock prediction algorithm. Firstly, we preprocessed the raw dataset and normalized data into the range from 0 to 1. Secondly, we introduced the LSTM model and improved its performance by tuning four parameters: learning rate, number of hidden layers, number of epochs, and batch size. Finally, we use four evaluation metrics to evaluate models: mean average error (MAE), root mean square error (RMSE), coefficient of determination (R2), and mean absolute error percentage (MAPE). Our LSTM model performs better than the previous model in experiments in terms of MAE, RMSE, R2, and MAPE.

An Adaptive Early Stopping Technique for DenseNet169-Based Knee Osteoarthritis Detection Model.

Knee osteoarthritis (OA) detection is an important area of research in health informatics that aims to improve the accuracy of diagnosing this debilitating condition. In this paper, we investigate the ability of DenseNet169, a deep convolutional neural network architecture, for knee osteoarthritis detection using X-ray images. We focus on the use of the DenseNet169 architecture and propose an adaptive early stopping technique that utilizes gradual cross-entropy loss estimation. The proposed approach allows for the efficient selection of the optimal number of training epochs, thus preventing overfitting. To achieve the goal of this study, the adaptive early stopping mechanism that observes the validation accuracy as a threshold was designed. Then, the gradual cross-entropy (GCE) loss estimation technique was developed and integrated to the epoch training mechanism. Both adaptive early stopping and GCE were incorporated into the DenseNet169 for the OA detection model. The performance of the model was measured using several metrics including accuracy, precision, and recall. The obtained results were compared with those obtained from the existing works. The comparison shows that the proposed model outperformed the existing solutions in terms of accuracy, precision, recall, and loss performance, which indicates that the adaptive early stopping coupled with GCE improved the ability of DenseNet169 to accurately detect knee OA.

0308 Impact of Customized Lighting on Sleep in Hospitalized Patients

Abstract Introduction Inpatient sleep may be disrupted for myriad reasons, including lighting exposure that is typically fixed at constant intensity and spectra. In this study we employed customized lighting that promotes circadian entrainment by exposing patients to blue-augmented lighting during the morning and blue-suppressed lighting evening to improve sleep timing and duration. Methods Two Mount Sinai hospital rooms (with 2 beds each) were outfitted with Eight Mattress sensors that use a combination of movement, breathing, heart rate, and temperature variables to estimate wake and sleep vigilance states in 30-second epochs. One room contained customized lighting between 7 am and 11 pm, while the other room contained standard hospital lighting. The current analysis includes 60 participants (30 in the control room and 30 in the custom lighting room), who were matched for time-of-year hospital admission date. We first analyzed the probability of sleep during any given 30-second epoch across each subject’s admission and compared the distributions of sleep probability across the 24-hour cycle for all subjects in the lighting vs control conditions. We additionally evaluated total nocturnal sleep duration(11 PM – 7 AM) by 1) summating all sleep epochs on nights containing at least 50% valid data and 2) normalizing all measured sleep epochs to all valid sleep/wake outputs (excluding indeterminate epochs). Results The distribution of sleep probability was significantly different in the lighting vs control condition (p < 0.001) with most apparent diversion of the distributions between 9 pm and 9 am. The average raw nocturnal sleep duration was significantly higher in lighting condition (270.0 ± 23 minutes vs 201 ± 23 min in control, p = 0.037), as was the average minutes of sleep per hour of valid recording (36.9 ± 15 minutes vs 28.4 ± 13 in control, p = 0.022). There were no significant differences in the number of nocturnal wake or indeterminate epochs between conditions. Conclusion Custom lighting appears to promote sleep in hospitalized patients, but the interpretation is tempered by absence of full validation of the mattress sensor against polysomnography. Future work will evaluate differences in sleep timing and continuity between conditions. Support (if any) McClung Foundation, R01 AG066870

Detection of Pneumonia from Chest X-ray Images Utilizing MobileNet Model.

Pneumonia has been directly responsible for a huge number of deaths all across the globe. Pneumonia shares visual features with other respiratory diseases, such as tuberculosis, which can make it difficult to distinguish between them. Moreover, there is significant variability in the way chest X-ray images are acquired and processed, which can impact the quality and consistency of the images. This can make it challenging to develop robust algorithms that can accurately identify pneumonia in all types of images. Hence, there is a need to develop robust, data-driven algorithms that are trained on large, high-quality datasets and validated using a range of imaging techniques and expert radiological analysis. In this research, a deep-learning-based model is demonstrated for differentiating between normal and severe cases of pneumonia. This complete proposed system has a total of eight pre-trained models, namely, ResNet50, ResNet152V2, DenseNet121, DenseNet201, Xception, VGG16, EfficientNet, and MobileNet. These eight pre-trained models were simulated on two datasets having 5856 images and 112,120 images of chest X-rays. The best accuracy is obtained on the MobileNet model with values of 94.23% and 93.75% on two different datasets. Key hyperparameters including batch sizes, number of epochs, and different optimizers have all been considered during comparative interpretation of these models to determine the most appropriate model.

PB3C-CNN: An integrated PB3C and CNN based approach for plant leaf classification

Plant identification and classification are critical to understand, protect, and conserve biodiversity. Traditional plant classification requires years of intensive training and experience, making it difficult for others to classify plants. Plant leaf classification is a challenging issue as similar features appears in different species of plant. With the development of automated image-based classification, machine learning (ML) is becoming very popular. Deep learning (DL) methods have significantly improved plant image identification and classification. In the last decade, convolutional neural networks (CNN) have entirely dominated the field of computer vision, showing outstanding feature extraction capabilities and significant identification and classification performance. The capability of CNN lies in its network. The primary strategy to continue this trend in the literature relies on further scaling networks in size. However, costs increase rapidly, while performance improvements may be marginal when the number of net-works increases. Hence, there is a need to optimize the CNN network to get the best possible result with the minimum number of networks and other parameters such as the number of epochs, number of layers, batch size and number of neurons. The paper aims to evolve the optimal architecture of CNN using PB3C algorithm for plant leaf classification. For this, we use the nature-inspired computing technique parallel big bang–big crunch to evolve a CNN's optimal architecture automatically. Current study validated the proposed approach for plant leaf classification and compared it with 11 other machine learning-based approaches. From the results obtained it was found that the proposed approach was able to outperforms all 11 existing state-of-the-art techniques.

Machine learning‐based regression models for predicting signal quality of dense wavelength division multiplexing (DWDM) optical communication network

SummaryOver the years, optical communication systems have been a significant source of fast and secure communication. However, factors like noise and mitigation error can degrade the bit error rate (BER) and quality factor (Q factor) of optical communication systems. Predicting the optimal threshold, Q factor, and BER is usually a difficult task. Therefore, in this paper, machine learning‐based linear regression, least absolute shrinkage and selection operator (LASSO) regression, and Ridge regression have been used for a dense wavelength division multiplexing (DWDM)‐based optical communication network to predict the signal quality. These techniques have been used to predict the desired BER, Q factor, threshold, and eye height of the system. To demonstrate this research concept, a DWDM‐based optical communication network of 50 km length is designed and simulated using Optisystem‐14.0. After data preparation, regression models have been developed and validated through diagnostic plots. Results show that mean square error (MSE) has a significant decline with an increase in the number of epochs for all four models. LASSO and Ridge regression have effectively resolved the issue of overfitting, which occurred in the linear regression case. Furthermore, the mean MSE plot proved the significant reduction of mean MSE in the case of LASSO regression. Results show that min BER for LASSO regression came out to be −173,627.14, providing a robust and cost‐efficient process.

Research of the Characteristics of a Convolutional Neural Network on the ESP32-CAM Microcontroller

The paper is devoted to solving the problem of using neural networks for real-time image recognition on low-power portable devices running on microcontrollers. The ESP-32© CAM microcontroller was used as the target device, on which an artificial neural network was deployed, written using the Python® programming language and the Tensorflow© library for building neural networks. The performance of the microcontroller and personal computer for object detection using a neural network and their classification were compared in the paper. The image recognition time and percentage of correctly classified objects were compared. The paper shows that the number of training epochs affects the accuracy of object classification in the image. The obtained results show that increasing the number of training epochs increases the accuracy of object recognition using the studied neural network, but a significant increase in the number of epochs does not lead to a significant improvement in recognition accuracy. The difference in the obtained results for the microcontroller and personal computer image recognition accuracy ranges from 5%.

Comparison of Different Parameters of Feedforward Backpropagation Neural Networks in DEM Height Estimation for Different Terrain Types and Point Distributions

Digital Elevation Models (DEMs) are commonly used for environment, engineering, and architecture-related studies. One of the most important factors for the accuracy of DEM generation is the process of spatial interpolation, which is used for estimating the height values of the grid cells. The use of machine learning methods, such as artificial neural networks for spatial interpolation, contributes to spatial interpolation with more accuracy. In this study, the performances of FBNN interpolation based on different parameters such as the number of hidden layers and neurons, epoch number, processing time, and training functions (gradient optimization algorithms) were compared, and the differences were evaluated statistically using an analysis of variance (ANOVA) test. This research offers significant insights into the optimization of neural network gradients, with a particular focus on spatial interpolation. The accuracy of the Levenberg–Marquardt training function was the best, whereas the most significantly different training functions, gradient descent backpropagation and gradient descent with momentum and adaptive learning rule backpropagation, were the worst. Thus, this study contributes to the investigation of parameter selection of ANN for spatial interpolation in DEM height estimation for different terrain types and point distributions.

The effect of artifact rejection on the performance of a convolutional neural network based algorithm for binary EEG data classification

ObjectiveEvaluate the effect of artifact rejection on the performance of a Convolutional Neural Network (CNN) based algorithm for classification of abnormal and normal electroencephalography (EEG) data. MethodsWe developed an automated CNN-based clean versus artifact classification algorithm and applied this to abnormal and normal EEG data. Additionally, algorithms for abnormal versus normal classification were developed with and without artifact rejection beforehand. For each algorithm, five CNNs were trained using 5-fold cross-validation and the majority vote of these was used for classification of test data. We compared bootstrap test accuracies and the number of training epochs required between the scenario in which artifacts were rejected and the scenario in which they were not. ResultsThe clean versus artifact classification algorithm had a test accuracy of 85% (recall:89%, precision:82%). Bootstrap test accuracies of the abnormal versus normal classification algorithms were similar with and without artifact rejection beforehand (mean [95% CI]:84% [77%-91%] and 84% [76%-91%], respectively). For the five cross-validation runs, respectively −8%, 22%, 31%, −94% and 26% less training epochs were required when artifacts were rejected beforehand. ConclusionIn our study, EEG artifact rejection did not improve CNN-based abnormal versus normal classification performance, but did slightly speed up training. SignificanceIn our study, artifact rejection was not required. The possibility to omit this step lowers the threshold for applying CNNs to EEG data and using these in clinical practice. Future studies should confirm whether this step could also be omitted in case other artificial intelligence techniques are applied to different datasets.

Learning Stress with Feet and Grids

This paper investigates quantity-insensitive stress learning using the MaxEnt learner of Pater and Prickett (2022) and compares the performance of the learner equipped with three different constraint sets: a foot-based constraint set and two grid-based constraint sets, one drawn directly from Gordon (2002), and one that changes the formulation of the main stress constraint to match the foot-based learner. The learner equipped with the foot-based constraint set succeeds at learning all the languages from the Gordon (2002) typology that it can represent; the structural ambiguity of the foot-based representations is not a problem in this regard. The foot-based learner also learns the languages as quickly in terms of number of epochs as the faster of the grid-based learners, which is the one with the revised main stress constraint. We conclude that the foot-based learner and the grid-based learner fare similarly well in this initial comparison on a typologically grounded set of learning problems.

Prediksi Indeks Harga Produsen Pertanian Karet Di Indonesia Menggunakan Metode LSTM

Natural rubber is one of the plantation commodities that has a fairly wide market in international trade because it is needed as a raw material for various industries. Rubber producer prices need to be predicted because producer prices are the first price in the lead from other price levels. So that information about price changes at the producer level is very important as an early warning system against price fluctuations at the next price level. The Long Short-Term Memory (LSTM) algorithm was chosen because it is considered capable of accommodating the problem of predicting the price index of producers in the rubber agriculture sector being faced because, LSTM itself is one of the developments of a neural network, which can be used for time series data modeling and is capable of continuous learning. . Parameter analysis carried out in this study is the number of hidden neurons, epochs and batch size. The best combination of parameters produced in this study is 50 hidden neurons, 25 epochs and batch size 10. The best values ​​generated in this study are the RMSE value of training data 384.20 and the value of RMSE testing 306.01 and the value of MAPE training 1.25% and the value of MAPE testing 1.09% The best MAPE error calculation in this study is "Predicting the Price Index of Agricultural Rubber Producers in Indonesia Using the Long Short Term Memory Method" which produces the best MAPE. These results indicate that the MAPE error can be said to be very good because the best MAPE value produced is below 10%.