Image classification is always a very important research topic in the medical field. The current medical image classification is mainly based on artificial intelligence methods. The main problem of existing methods is...
详细信息
ISBN:
(纸本)9798350377620;9798350377613
Image classification is always a very important research topic in the medical field. The current medical image classification is mainly based on artificial intelligence methods. The main problem of existing methods is the slow speed, precision of training, and redundant iteration. To solve this problem, Li et al. proposed Gaussian process regression-based learning rate optimization (GLRO) method. This method solves the problem of redundant iteration and improves classification accuracy. Gaussian process regression is used to improve the optimization effect of the optimizer and further improve the classification accuracy. However, the limitation of this method is that it uses a single training network and loss function. To improve this medtod, we propose a more efficient federated learning method called FedGPM. We train on many classification networks with loss functions more appropriate to the properties of medical images. The results show that our algorithm improves on the original GLRO algorithm and has a very good performance. The accuracy rate on the NIH chest X-ray dataset is 95% to 97%, and the accuracy rate on the specific dataset is close to 99%.
This research proposes a series of novel learning rate optimization algorithms with two versions for Adaptive Moment Estimation (Adam), which is a common optimizer in Convolutional Neural Networks (CNNs). Optimizer th...
详细信息
This research proposes a series of novel learning rate optimization algorithms with two versions for Adaptive Moment Estimation (Adam), which is a common optimizer in Convolutional Neural Networks (CNNs). Optimizer that is used to control the training efficiency and prediction accuracy by controlling the convergence progress plays an important role in CNNs. However, optimizers such as Adam are usually not hyperparameter-free and very sensitive to the hyperparameters embedded in CNNs. For example, the learningrate is a hyperparameter that represents a step size in the calculations. The learningrate has the most significant influence on prediction accuracy, so optimizing the learningrate is the best way to improve accuracy. In this research, a series of Gaussian Process Regression (GPR)-based learning rate optimization (GLRO) algorithms are proposed to increase the classification accuracy. To be specific, the relationship between the learningrate and corresponding accuracy is studied and the potential learningrate is predicted by the GPR model which is built with previous learningrates and corresponding accuracies. Also, two strategies of the algorithm to select the input learningrate are tested separately. AlexNet, which is a state-of-the-art CNN, is used as a framework to evaluate the proposed algorithms. AlexNet is widely used in the healthcare system as medical imaging classification framework. The Stimulated Raman Scattering (SRS) images of human brain tumors are used to classify cells and non-cells in this research. The proposed GLRO are compared to the conventional learningrate annealing algorithm and the constant learningrate algorithm. The algorithms' classifications of SRS images are evaluated in terms of accuracy, sensitivity, specificity, and precision. To further validate GLRO, multiple benchmark medical images and CNN frameworks are tested. The experimental results illustrate that the proposed GLRO algorithms outperform other algorithms by showi
Purpose: This study aimed to enhance the prediction of container dwell time, a crucial factor for optimizing port operations, resource allocation, and supply chain efficiency. Determining an optimal learningrate for ...
详细信息
Purpose: This study aimed to enhance the prediction of container dwell time, a crucial factor for optimizing port operations, resource allocation, and supply chain efficiency. Determining an optimal learningrate for training Artificial Neural Networks (ANNs) has remained a challenging task due to the diverse sizes, complexity, and types of data involved. Design/Method/Approach: This research used a RandomizedSearchCV algorithm, a random search approach, to bridge this knowledge gap. The algorithm was applied to container dwell time data from the TOS system of the Port of Tema, which included 307,594 container records from 2014 to 2022. Findings: The RandomizedSearchCV method outperformed standard training methods both in terms of reducing training time and improving prediction accuracy, highlighting the significant role of the constant learningrate as a hyperparameter. Research Limitations and Implications: Although the study provides promising outcomes, the results are limited to the data extracted from the Port of Tema and may differ in other contexts. Further research is needed to generalize these findings across various port systems. Originality/Value: This research underscores the potential of RandomizedSearchCV as a valuable tool for optimizing ANN training in container dwell time prediction. It also accentuates the significance of automated learningrate selection, offering novel insights into the optimization of container dwell time prediction, with implications for improving port efficiency and supply chain operations.
An improved differential evolution (IDE) algorithm-based Elman neural network (ENN) controller is proposed to control a squirrel-cage induction generator (SCIG) system for grid-connected wind power applications. First...
详细信息
An improved differential evolution (IDE) algorithm-based Elman neural network (ENN) controller is proposed to control a squirrel-cage induction generator (SCIG) system for grid-connected wind power applications. First, the control characteristics of a wind turbine emulator are introduced. Then, an AC/DC converter and a DC/AC inverter are developed to convert the electric power generated by a three-phase SCIG to the grid. Moreover, the dynamic model of the SCIG system is derived for the control of the square of DC-link voltage according to the principle of power balance. Furthermore, in order to improve the transient and steady-state responses of the square of DC-link voltage of the SCIG system, an IDE-based ENN controller is proposed for the control of the SCIG system. In addition, the network structure and the online learning algorithm of the ENN are described in detail. Additionally, according to the different wind speed variations, a lookup table built offline by the dynamic model of the SCIG system using the IDE is provided for the optimisation of the learningrates of ENN. Finally, to verify the control performance, some experimental results are provided to verify the feasibility and the effectiveness of the proposed SCIG system for grid-connected wind power applications.
暂无评论