Academic literature on the topic 'Kaiming initialization'

CHD is chest pain or discomfort that occurs if the area of the heart muscle does not get enough oxygen-rich blood. CHD is also known as coronary artery disease. CHD is increasing every year with a significant number of deaths. A learning algorithm is proposed to get better performance in accuracy, sensitivity, and specificity in CHD interpretation. Accuracy can be improved by adding a Kaiming He (2015) weight initialization optimization technique to the DNN structure. Therefore we propose that DNN is optimized with a Kaiming He weight initialization technique so that it can overcome weaknesses in the data variant. This is evidenced by the results of the accuracy performance of 98.73%. Initialization of kaiming he weights is proven to improve accuracy and overcome the problem of large data variants between classes

<p>In the scenario of 3C (Computer, Communication, Consumer Electronics), the algorithm for detecting targets in smartphone component assembly consumes a substantial amount of system computing resources.It also faces challenges such as the flexible nature of target components and the small scale of heterogeneous components, leading to low detection accuracy. To adapt to the 3C scenario, this paper proposes improvements based on the DINO object detection model. It introduces a more lightweight and powerful feature extraction backbone, Efficientnetv2, and utilizes the He-Kaiming weight initialization method to extract strong multi-scale feature maps. In training, a more efficient dynamic contrastive denoising training method is employed. This approach makes the model lightweight and accurate for 3C detection. This method outperforms leading detection algorithms in both accuracy of experimental results and parameter efficiency.</p> <p> </p>

In the field of audio classification, speech recognition or environmental sound classification, the performance has been greatly improved using deep learning based systems, but is still a challenge where comes small scale corpora training as the deep learning based systems need a huge amount of training data and it is not easy to get that much data. For this problem, this paper proposed a solution of multi-augmented data classification using convolutional neural network with kaiming initialization for limited resources (MA-CNN). The contributions of this work are twofold. First we propose a 2-dimension convolutional neural network with kaiming initialization, using only convolutional and fully connected layers. This prevent the neural network from exploding in the forward pass process. Secondly, to avoid data scarcity, over-fitting and to improve model robustness, we used multiple data augmentation techniques that increased training data quantity. The proposed methodology outperformed CNN without data augmentation technique.

Abstract In this work, we generalize the ideas of Kaiming initialization to Graph Neural Networks (GNNs) and propose a new scheme (G-Init) that reduces oversmoothing, leading to very good results in node and graph classification tasks. GNNs are commonly initialized using methods designed for other types of Neural Networks, overlooking the underlying graph topology. We analyze theoretically the variance of signals flowing forward and gradients flowing backward in the class of convolutional GNNs. We then simplify our analysis to the case of the GCN and propose a new initialization method. Results indicate that the new method (G-Init) reduces oversmoothing in deep GNNs, facilitating their effective use. Our approach achieves an accuracy of 61.60% on the CS dataset (32-layer GCN) and 69.24% on Cora (64-layer GCN), surpassing state-of-the-art initialization methods by 25.6 and 8.6 percentage points, respectively. Extensive experiments confirm the robustness of our method across multiple benchmark datasets, highlighting its effectiveness in diverse settings. Furthermore, our experimental results support the theoretical findings, demonstrating the advantages of deep networks in scenarios with no feature information for unlabeled nodes (i.e., “cold start” scenario).

Summary Hydraulic fracturing is an effective method for enhancing both the initial reservoir production and ultimate recovery. Nevertheless, the conductivity of proppant fractures is a pivotal factor in the optimization of fracture designs within the context of fracture modification. Experimental testing methods for proppant fracture conductivity are costly and time-consuming, and the physical model is excessively complex and incomplete to account for all the influencing factors, resulting in low computational efficiency. A backpropagation neural network (BPNN) model was constructed using the RAdam optimization algorithm to identify a more efficacious method for predicting the conductivity of proppant fractures. The model was used to predict the fracture conductivity of two data types pertaining to the experimental data on the conductivity of geothermal and volcanic reservoirs. The prediction model is enhanced for three key areas. First, an isolated forest algorithm is used to assess and discard anomalous data points. Second, the objective function is optimized by employing the RAdam optimization algorithm, which has the advantages of both Adam and stochastic gradient descent (SGD). This guarantees rapid convergence and prevents the initial training phase from converging to a locally optimal solution. Moreover, this approach enhances the stability of the model training process. Finally, the rectified linear unit (ReLU) activation function may result in issues related to neuronal activity, including the potential for its disappearance. This study addresses this problem by employing the Kaiming initialization method. The experiments used a series of evaluation metrics, including the mean square error and coefficient of determination, to assess the predictive performance of the two data sets in the two distinct models. The experimental results indicate that the BPNN with RAdam optimization is a more effective approach for data pertaining to volcanic and geothermal reservoirs. Moreover, the prediction of the geothermal reservoir data is more precise than that of the volcanic reservoir data. This model can be used for rapid predictions based on existing fracture conductivity data, which can better guide the design of fracture modifications and is of great importance.