Network Model Research Articles

Pineapple Detection with YOLOv7-Tiny Network Model Improved via Pruning and a Lightweight Backbone Sub-Network

High-complexity network models are challenging to execute on agricultural robots with limited computing capabilities in a large-scale pineapple planting environment in real time. Traditional module replacement often struggles to reduce model complexity while maintaining stable network accuracy effectively. This paper investigates a pineapple detection framework with a YOLOv7-tiny model improved via pruning and a lightweight backbone sub-network (the RGDP-YOLOv7-tiny model). The ReXNet network is designed to significantly reduce the number of parameters in the YOLOv7-tiny backbone network layer during the group-level pruning process. Meanwhile, to enhance the efficacy of the lightweight network, a GSConv network has been developed and integrated into the neck network, to further diminish the number of parameters. In addition, the detection network incorporates a decoupled head network aimed at separating the tasks of classification and localization, which can enhance the model’s convergence speed. The experimental results indicate that the network before pruning optimization achieved an improvement of 3.0% and 2.2%, in terms of mean average precision and F1 score, respectively. After pruning optimization, the RGDP-YOLOv7-tiny network was compressed to just 2.27 M in parameter count, 4.5 × 109 in computational complexity, and 5.0MB in model size, which were 37.8%, 34.1%, and 40.7% of the original YOLOv7-tiny network, respectively. Concurrently, the mean average precision and F1 score reached 87.9% and 87.4%, respectively, with increases of 0.8% and 1.3%. Ultimately, the model’s generalization performance was validated through heatmap visualization experiments. Overall, the proposed pineapple object detection framework can effectively enhance detection accuracy. In a large-scale fruit cultivation environment, especially under the constraints of hardware limitations and limited computational power in the real-time detection processes of agricultural robots, it facilitates the practical application of artificial intelligence algorithms in agricultural engineering.

Research on high-frequency trading strategies optimization based on deep learning

High-frequency trading is derived from programmatic trading and market maker mechanisms, and unlike low-frequency trading, it uses ultra-high-speed computers to acquire and analyze high-frequency trading data in the market, so as to identify price change patterns and quickly execute trades to complete the change of hands. In this work, we focused on optimizing high-frequency trading strategies using deep reinforcement learning. Specifically, we employ a dual deep Q network model, which consists of two deep neural networks that are used to generate action value estimates and provide target action values, respectively. The advantage of this structure is that it reduces bias in the estimation process, thereby improving the stability and efficiency of learning. The input layer of the model receives multi-dimensional features of market data, such as price, volume, order depth, etc., which are normalized and then fed into the network. The hidden layer includes multiple layers of fully connected layers to enhance the nonlinear representation of the model and the ability to handle complex market dynamics. We have also introduced batch normalization and dropout layers in the network to prevent overfitting merges and improve the generalization ability of the model. Eventually, the output layer generates action values corresponding to each possible trading action, deciding to buy, sell, or hold a position. In the experimental analysis, this study verifies the superiority of the proposed model in processing high-frequency trading data and executing trading decisions by comparing it with traditional trading strategies and other machine learning methods.

Geometric Optimization Algorithm for Path Loss Model of Riparian Zone IoT Networks Based on Federated Learning Framework

Geometric Optimization Algorithm for Path Loss Model of Riparian Zone IoT Networks Based on Federated Learning Framework

Soil temperature prediction based on explainable artificial intelligence and LSTM

Soil temperature is a key parameter in many disciplines, and its research has important practical significance. In recent years, the prediction of soil temperature by deep learning has achieved good results. However, deep learning is difficult to popularize in practical use because of its opacity. This study aims to interpret and analyze the Long Short Term Memory Network (LSTM) model for global soil temperature prediction using SHapley Additive exPlanation (SHAP), Permutation Importance (PI) and Partial Dependence Plot (PDP). The results show that Temperature of air at 2 m above the surface of land has the greatest influence on the prediction of soil temperature, and its SHAP and PI characteristic values have significant seasonality. Meanwhile, radiation also has a certain influence on the prediction results. There was a significant positive correlation between the temperature of 2 m and the soil temperature. The explanatory insights provided in this paper enhance the transparency and confidence of the model, which promotes the applicability of soil temperature prediction models in relevant fields.

Mean Squared Error May Lead You Astray When Optimizing Your Inverse Design Methods

Abstract When performing time-intensive optimization tasks, such as those in Topology or Shape Optimization, researchers have turned to machine-learned (ML) Inverse Design methods i.e., predicting the optimized geometry from input conditions to replace or warm start traditional optimizers. Such methods are often optimized to reduce the Mean Squared Error (MSE) or Binary Cross-Entropy between the output and a training dataset of optimized designs. While convenient, we show that this choice may be myopic. Specifically, we compare two methods of optimizing the hyper-parameters of easily reproducible Machine Learning models including random forest (RF), k-nearest neighbors (KNN), and Deconvolutional Neural Network (DeCNN) model for predicting the three optimal topology problems. We show that under both direct Inverse Design and when warm starting further Topology Optimization (TO), using MSE metrics to tune hyper-parameters produces less performant models than directly evaluating the objective function, though both produce designs that are almost one order of magnitude better than using the common uniform initialization. We also illustrate how warm starting impacts both the convergence time, the type of solutions obtained during optimization, and the final designs. Overall, our initial results portend that researchers may need to revisit common choices for evaluating ID methods that subtly trade-off factors in how an ID method will actually be used. We hope our open-source dataset and evaluation environment will spur additional research in those directions.

Study on hysteresis performance of four-limb CFST latticed column-box girder joints based on GA-BP neural network

This study investigates the seismic performance and energy dissipation capacity of the concrete-filled steel tube (CFST) latticed column-box girder joints. A low-cycle reciprocating load test was carried out on the CFST latticed column-box girder joints with three different connection forms. Based on the test results, a genetic algorithm (GA) is developed in combination with the back propagation neural network (BPNN) model for prediction research. Using a training set of joints with single-limb diagonal and cross-diagonal braces (training set), the development of load-displacement hysteresis curves of joints with transverse diaphragm plate (test set) under low-cycle reciprocating loads is predicted. The test results are compared with those obtained from the traditional algorithms and finite element simulations. The results on the joint hysteresis performance show that all three types of joints have a good seismic performance and energy dissipation capacity. Compared with the traditional algorithms, the GA-BP model predicts more accurate results and is therefore used to predict the load-displacement hysteresis curves of CFST latticed column-box girder joints, which is shown to be convenient and feasible.

THE EFFECT OF ANTHROPOMETRIC PARAMETERS ON THE SEAT TRANSMISSIBILITY DURING WHOLE-BODY VIBRATION: A HYBRID MODELING APPROACH BASED ON ARTIFICIAL NEURAL NETWORKS AND PRINCIPAL COMPONENT ANALYSIS

The modeling of the seat transmissibility is necessary to advance the understanding of the dynamic interactions between compliant seats and occupants. Within this investigation, an optimized artificial neural network (ANN) model was employed to clarify contributions associated with anthropometric parameters to the seat transmissibility. Anthropometric parameters underwent dimensionality reduction through the principal component analysis, and resultant principal components served as input features for the ANN model. Additionally, the ANN structure’s weights and biases values were adjusted using the genetic algorithm (GA), resulting in a PCA-GA-ANN model for the prediction of seat transmissibilities. The results indicated root mean square error (RMSE) values for predicting vertical in-line and horizontal cross-axis transmissibilities from the developed model were 0.061 and 0.055, respectively, demonstrating superior effectiveness in the prediction error and trends when compared with both the ANN and GA-ANN models. The seat transmissibility predicted from the PCA-GA-ANN model exhibited resonance behaviors similar to that observed in the whole-body vibration test. The sensitivity analysis showed that the subject’s age was the most predominant anthropometric parameter for the prediction, followed by gender and body mass index. The ANN model optimized with principal component analysis (PCA) and GA effectively eliminates the redundant information of anthropometric parameters, enhancing the generalization of the seat transmissibility prediction.

An Intelligent Vehicle Price Estimation Approach Using a Deep Neural Network Model

In recent years, the market for used-vehicle trade in the Kingdom of Saudi Arabia has grown significantly. This is due to the high cost of new vehicles that are not affordable by most buyers and lifting the ban on women drivers. Recently, several online websites for selling vehicles are available with different functions. However, estimating the vehicle price is based on traditional calculation methods, and this is inaccurate in several selling situations, as there are many factors that may affect the vehicle price, and these factors must be taken into consideration when estimating the vehicle’s price. Therefore, there is high demand to develop an automated vehicle price estimation system through adopting artificial intelligence (AI) technologies. Hence, this paper proposes an efficient vehicle price estimation system through developing an efficient deep neural network (DNN) model. The developed DNN model has been trained using a recent collected dataset for used-vehicle prices in the Kingdom of Saudi Arabia. The developed system has been validated using a recent vehicle price dataset, and the obtained results are compared with seven different machine learning models and showed a promising regression accuracy. In addition, we developed a reliable graphical user interface (GUI) for the purpose of allowing the user to estimate the price of any vehicle using the pre-trained DNN model.

Attention-enhanced dilated convolution for Parkinson’s disease detection using transcranial sonography

BackgroundTranscranial sonography (TCS) plays a crucial role in diagnosing Parkinson's disease. However, the intricate nature of TCS pathological features, the lack of consistent diagnostic criteria, and the dependence on physicians' expertise can hinder accurate diagnosis. Current TCS-based diagnostic methods, which rely on machine learning, often involve complex feature engineering and may struggle to capture deep image features. While deep learning offers advantages in image processing, it has not been tailored to address specific TCS and movement disorder considerations. Consequently, there is a scarcity of research on deep learning algorithms for TCS-based PD diagnosis.MethodsThis study introduces a deep learning residual network model, augmented with attention mechanisms and multi-scale feature extraction, termed AMSNet, to assist in accurate diagnosis. Initially, a multi-scale feature extraction module is implemented to robustly handle the irregular morphological features and significant area information present in TCS images. This module effectively mitigates the effects of artifacts and noise. When combined with a convolutional attention module, it enhances the model's ability to learn features of lesion areas. Subsequently, a residual network architecture, integrated with channel attention, is utilized to capture hierarchical and detailed textures within the images, further enhancing the model's feature representation capabilities.ResultsThe study compiled TCS images and personal data from 1109 participants. Experiments conducted on this dataset demonstrated that AMSNet achieved remarkable classification accuracy (92.79%), precision (95.42%), and specificity (93.1%). It surpassed the performance of previously employed machine learning algorithms in this domain, as well as current general-purpose deep learning models.ConclusionThe AMSNet proposed in this study deviates from traditional machine learning approaches that necessitate intricate feature engineering. It is capable of automatically extracting and learning deep pathological features, and has the capacity to comprehend and articulate complex data. This underscores the substantial potential of deep learning methods in the application of TCS images for the diagnosis of movement disorders.Graphical

Neural waves and computation in a neural net model II: Data-like structures and the dynamics of episodic memory.

The computational resources of a neuromorphic network model introduced earlier were investigated in the first paper of this series. It was argued that a form of ubiquitous spontaneous local convolution enabled logical gate-like neural motifs to form into hierarchical feed-forward structures of the Hubel-Wiesel type. Here we investigate concomitant data-like structures and their dynamic rôle in memory formation, retrieval, and replay. The mechanisms give rise to the need for general inhibitory sculpting, and the simulation of the replay of episodic memories, well known in humans and recently observed in rats. Other consequences include explanations of such findings as the directional flows of neural waves in memory formation and retrieval, visual anomalies and memory deficits in schizophrenia, and the operation of GABA agonist drugs in suppressing episodic memories. We put forward the hypothesis that all neural logical operations and feature extractions are of the convolutional hierarchical type described here and in the earlier paper, and exemplified by the Hubel-Wiesel model of the visual cortex, but that in more general cases the precise geometric layering might be obscured and so far undetected.

Bipolar disorder: Construction and analysis of a joint diagnostic model using random forest and feedforward neural networks

BackgroundTo construct a diagnostic model for Bipolar Disorder (BD) depressive phase using peripheral tissue RNA data from patients and combining Random Forest with Feedforward Neural Network methods. MethodsDatasets GSE23848, GSE39653, and GSE69486 were selected, and differential gene expression analysis was conducted using the limma package in R. Key genes from the differentially expressed genes were identified using the Random Forest method. These key genes' expression levels in each sample were used to train a Feedforward Neural Network model. Techniques like L1 regularization, early stopping, and dropout layers were employed to prevent model overfitting. Model performance was then validated, followed by GO, KEGG, and protein-protein interaction network analyses. ResultsThe final model was a Feedforward Neural Network with two hidden layers and two dropout layers, comprising 2345 trainable parameters. Model performance on the validation set, assessed through 1000 bootstrap resampling iterations, demonstrated a specificity of 0.769 (95 % CI 0.571–1.000), sensitivity of 0.818 (95 % CI 0.533–1.000), AUC value of 0.832 (95 % CI 0.642–0.979), and accuracy of 0.792 (95 % CI 0.625–0.958). Enrichment analysis of key genes indicated no significant enrichment in any known pathways. ConclusionKey genes with biological significance were identified based on the decrease in Gini coefficient within the Random Forest model. The combined use of Random Forest and Feedforward Neural Network to establish a diagnostic model showed good classification performance in Bipolar Disorder.

A network analysis of body image concern, interoceptive sensibility, self-consciousness, and self-objectification.

Disrupted interoception has been found in individuals with body image concerns and related psychiatric conditions (i.e., eating disorders, body dysmorphic disorder). However, few studies have explored the role of interoception in the relationship between body image concern and self-objectification (SO). As components of the bodily self, interoception and body image may also contribute to the fostering of self-consciousness. The current study estimated the network model of body image concern, adaptive and maladaptive interoceptive sensibility (IS), self-consciousness, and SO, with data from 743 neurotypical adults. The centrality of nodes and the robustness of the network were verified. Our network showed that interoception correlated with certain characteristics of body image concern (BIC) with varying strength and different directions. The most central node described beliefs of BIC, namely, negative evaluation toward appearance. Self-objectification and maladaptive IS presented the highest bridge centrality. Interoception also linked BICs with SO and self-consciousness. Findings suggested that SO, adaptive and maladaptive interoception might be the targets of intervention for individuals with BIC, specifically for those with high levels of appearance-related negative self-evaluations.

Contrasting bean analysis system based on YOLOv5 and a neural network model using the interval type-2 fuzzy set approach

Contrasting bean analysis system based on YOLOv5 and a neural network model using the interval type-2 fuzzy set approach

Artificial Neural Network Model to Predict OBrix and pH of Banana Based on Color Parameters

Artificial neural network (ANN) was used to predict internal quality parameters (oBrix and pH) of lady finger banana. This research consisted of three stages, namely: (1) capturing images of lady finger banana using a computer vision system; (2) measurement of oBrix and pH of the banana; (3) ANN architecture analysis using the Matlab R2019a application. The ANN architectural model consisted of 3 output models, namely: (1) oBrix values; (2) pH value; (3) oBrix and pH values. The ANN architecture analysis was carried out through two phases. Phase I consisted of 45 experimental units and phase II with 35 experimental units. The best ANN architecture to be used as a prediction model for oBrix and pH of golden banana fruit is ANN architecture model 3 with the number of neurons inside the hidden layer = 3; activation function in hidden layer = logsig; activation function inside the output layer = logsig; data transformation range 0 – 1; learning rate value = 0.01; learning algorithm = tradingda; with MSE (mean square error), MAE (mean absolute error) performance and R correlation coefficient from training results of 0.0954; 0.2619 and 0.6538; test results 0.0392; 0.1606 and 0.7000 and validation results 0.0289; 0.1474 and 0.7889. Keywords: Artificial neural networks, Color, Computer vision system, CVS, RGB.

Analyzing risk influencing factors of ship collision accidents: A data-driven Bayesian network model integrating physical knowledge

The interaction and propagation effects among risk-influencing factors (RIFs) often lead to maritime accidents. Therefore, to support maritime risk management in decision-making, it is imperative to conduct a quantitative risk assessment (QRA) of these accidents, which requires a methodology capable of capturing causal relationships in limited cases. To achieve this goal, this study proposes a data-driven Bayesian network (BN) model that integrates physical knowledge with the QRA. Based on the collected data, a combination of domain knowledge, structure learning, and parameter learning is employed to construct the model. In the data-driven phase, three structural learning algorithms including Bayesian search (BS), Greedy thick thinning (GTT), and Peter-Clark (PC) algorithms, were used in combination with the expectation maximization algorithm of parameter learning. Then, using four indices calculated by the confusion matrix, the most fitting algorithm for structure learning was chosen, and the confusion matrix was obtained by five-fold cross-validation. Finally, network propagation impact (NPI) is introduced to prioritize RIFs. Preventive measures and emergency plans are formulated based on the network resilience metric (NRM) and network vulnerability metric (NVM). This approach leverages the objectivity of a data-driven BN and integrates domain expertise to enhance model validity. A case study of collisions was conducted to demonstrate the applicability of the model. The data were sourced from 327 collision accident reports provided by the China Maritime Safety Administration, and the PC algorithm was chosen. Findings indicate that the “Management system”, “Supervision”, and “Crew training” RIFs are prioritized for preventive measures, whereas “Communication” receives more resources for emergency handling. The results suggest the formulation of risk management strategies for practitioners to improve maritime safety.

Implementation of Convolutional Neural Network Algorithms to Detect the Ripeness of Palm Fruits based on Image Colors

The ripeness of oil palm fruit greatly affects the quality of the oil produced. Usually, humans manually detect the maturity of oil palm fruit, a process that can be inconsistent and time-consuming. Therefore, there is a need for automated methods that can improve accuracy and efficiency in determining the maturity of oil palm fruit. The study proposes using the Convolutional Neural Network algorithm to detect oil palm fruit maturity based on digital color images. We expect a convolutional neural network, known for its effectiveness in image analysis, to deliver accurate results in classifying fruit maturity. The aim of this study is to develop and test a model of a convolutional neural network that can classify oil palm fruit into three maturity categories: raw, ripe, and rotten. This study employs a maturity-categorized data set of oil palm images. We built and trained a convolutional neural network model using this data set. We evaluate the model performance using four main metrics: accuracy, precision, recall, and f1-score. We perform an in-depth analysis to assess model performance in each maturity category. The confusion matrix yielded an accuracy of 74.6%. The convolutional neural network model developed showed the highest precision results for the ripe fruit (86.00%), followed by rotten fruit (72.30%), and raw fruit (67.30%). We obtained the highest recalls (85.40%) for the raw fruit category, followed by ripe fruits (75.50%) and rotten fruits (72.30%.

Hybrid PDES Simulation of HPC Networks Using Zombie Packets

Although high-fidelity network simulations have proven to be reliable and cost-effective tools to peer into architectural questions for high-performance computing (HPC) networks, they incur a high resource cost. The time spent in simulating a single millisecond of network traffic in the highest detail can take hours, even for static, well-behaved traffic patterns such as uniform random. Surrogate models offer a significant reduction in runtime, yet they cannot serve as complete replacements and should only be used when appropriate. Thus, there is a need for hybrid modeling, where high-fidelity simulation and surrogates run side-by-side. We present a surrogate model for HPC networks in which: packets bypass the network, while the network state is left untouched, i.e , suspended. To bypass the network, we use historical data to estimate the arrival time at which every packet should be scheduled at; to suspend the network, all in-flight packets are scheduled to arrive at their destinations, and are kept in the system to awaken as zombies when switching back to high-fidelity. Speedup for a hybrid model is relative to the proportion of surrogate to high-fidelity. This light-weight surrogate obtained up to 76 × speedup. Keeping the zombies in the network showed an increase in the accuracy of the high-fidelity simulation on restart when compared to restarting the network from an empty state.

Prediction of the fundamental viscoelasticity of asphalt mixtures using ML algorithms

The complex modulus (E*) and phase angle (δ) are fundamental viscoelastic (VE) properties of asphalt material, crucial for pavement performance analysis and modeling. This study presents a novel approach to predict the fundamental VE properties of asphalt mixtures using 20 study mixes with disparate mix design information including different Performance Graded (PG grade) binders, recycled asphalt (RAP) contents, with/without polymer modifications, and mix volumetric properties. By employing the XGBoost model for feature importance analysis, integrating the Bailey method to streamline gradation information, and utilizing master curve shape parameters for modeling, this study has effectively reduced data redundancy and dimensionality. Among four machine learning (ML) models compared, the Multi-Hidden-Layer Backpropagation Neural Network (Multi BP) model demonstrates superior applicability to the asphalt mixture dataset, providing accurate predictions for VE properties at any temperature and frequency combinations. Notably, this predictive model relies solely on the mixture design information as inputs, minimizing the need for laborious laboratory experiments. The developed model can be integrated with Mechanistic-Empirical (ME) design, predicting the variations in rutting and cracking performance of asphalt mixtures over the entire service life. This provides essential foundational input support for long-lasting pavement design.

Automated detection of tonic seizures using wearable movement sensor and artificial neural network.

Although several validated wearable devices are available for detection of generalized tonic-clonic seizures, automated detection of tonic seizures is still a challenge. In this phase 1 study, we report development and validation of an artificial neural network (ANN) model for automated detection of tonic seizures with visible clinical manifestation using a wearable wristband movement sensor (accelerometer and gyroscope). The dataset prospectively recorded for this study included 70 tonic seizures from 15 patients (seven males, age 3-46 years, median = 19 years). We trained an ANN model to detect tonic seizures. The independent test dataset comprised nocturnal recordings, including 10 tonic seizures from three patients and additional (distractor) data from three subjects without seizures. The ANN model detected nocturnal tonic seizures with visible clinical manifestation with a sensitivity of 100% (95% confidence interval = 69%-100%) and with an average false alarm rate of .16/night. The mean detection latency was 14.1 s (median = 10 s), with a maximum of 47 s. These data suggest that nocturnal tonic seizures can be reliably detected with movement sensors using ANN. Large-scale, multicenter prospective (phase 3) trials are needed to provide compelling evidence for the clinical utility of this device and detection algorithm.

An intelligent model for efficient load forecasting and sustainable energy management in sustainable microgrids

Microgrids have emerged as a promising solution for enhancing energy sustainability and resilience in localized energy distribution systems. Efficient energy management and accurate load forecasting are one of the critical aspects for improving the operation of microgrids. Various approaches for energy prediction and load forecasting using statistical models are discussed in the literature. In this work, a novel energy management framework that incorporates machine learning (ML) techniques is presented for an accurate prediction of solar and wind energy generation. The anticipated approach also emphasizes time series-based load forecasting in microgrids with precise estimation of State of Charge (SoC) of battery. A unique feature of the proposed framework is that utilizes historical load data and employs time series analysis coupled with different ML models to forecast the load demand in a commercial microgrids scenario. In this work, Long Short-Term Memory (LSTM) and Linear Regression (LR) models are employed for an experimental analysis to study the proposed framework under three different cases, such as (i) prediction of energy generation, (ii) load demand forecasting and, (iii) prediction of SoC of battery. The results show that the Random Forest (RF) and LSTM models performs well for energy prediction and load forecasting respectively. On the other hand, the Artificial Neural Network (ANN) model exhibited superior accuracy in terms of SoC estimation. Further, in this work, a Graphical User Interface (GUI) is developed for evaluating the efficacy of the proposed energy management framework.