RESUMEN
BACKGROUND AND PURPOSE: The immunohistochemical test (IHC) for Human Epidermal Growth Factor Receptor 2 (HER2) and hormone receptors (HR) provides prognostic information and guides treatment for patients with invasive breast cancer. The objective of this paper is to establish a non-invasive system for identifying HER2 and HR in breast cancer using dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI). METHODS: In light of the absence of high-performance algorithms and external validation in previously published methods, this study utilizes 3D deep features and radiomics features to represent the information of the Region of Interest (ROI). A Siamese Neural Network was employed as the classifier, with 3D deep features and radiomics features serving as the network input. To neutralize manufacturer bias, a batch effect normalization method, ComBat, was introduced. To enhance the reliability of the study, two datasets, Predict Your Therapeutic Response with Imaging and moLecular Analysis (I-SPY 1) and I-SPY 2, were incorporated. I-SPY 2 was utilized for model training and validation, while I-SPY 1 was exclusively employed for external validation. Additionally, a breast tumor segmentation network was trained to improve radiomic feature extraction. RESULTS: The results indicate that our approach achieved an average Area Under the Curve (AUC) of 0.632, with a Standard Error of the Mean (SEM) of 0.042 for HER2 prediction in the I-SPY 2 dataset. For HR prediction, our method attained an AUC of 0.635 (SEM 0.041), surpassing other published methods in the AUC metric. Moreover, the proposed method yielded competitive results in other metrics. In external validation using the I-SPY 1 dataset, our approach achieved an AUC of 0.567 (SEM 0.032) for HR prediction and 0.563 (SEM 0.033) for HER2 prediction. CONCLUSION: This study proposes a non-invasive system for identifying HER2 and HR in breast cancer. Although the results do not conclusively demonstrate superiority in both tasks, they indicate that the proposed method achieved good performance and is a competitive classifier compared to other reference methods. Ablation studies demonstrate that both radiomics features and deep features for the Siamese Neural Network are beneficial for the model. The introduced manufacturer bias normalization method has been shown to enhance the method's performance. Furthermore, the external validation of the method enhances the reliability of this research. Source code, pre-trained segmentation network, Radiomics and deep features, data for statistical analysis, and Supporting Information of this article are online at: https://github.com/FORRESTHUACHEN/Siamese_Neural_Network_based_Brest_cancer_Radiogenomic.
RESUMEN
BACKGROUND: Noonan syndrome (NS) is a rare genetic disease, and patients who suffer from it exhibit a facial morphology that is characterized by a high forehead, hypertelorism, ptosis, inner epicanthal folds, down-slanting palpebral fissures, a highly arched palate, a round nasal tip, and posteriorly rotated ears. Facial analysis technology has recently been applied to identify many genetic syndromes (GSs). However, few studies have investigated the identification of NS based on the facial features of the subjects. OBJECTIVES: This study develops advanced models to enhance the accuracy of diagnosis of NS. METHODS: A total of 1,892 people were enrolled in this study, including 233 patients with NS, 863 patients with other GSs, and 796 healthy children. We took one to 10 frontal photos of each subject to build a dataset, and then applied the multi-task convolutional neural network (MTCNN) for data pre-processing to generate standardized outputs with five crucial facial landmarks. The ImageNet dataset was used to pre-train the network so that it could capture generalizable features and minimize data wastage. We subsequently constructed seven models for facial identification based on the VGG16, VGG19, VGG16-BN, VGG19-BN, ResNet50, MobileNet-V2, and squeeze-and-excitation network (SENet) architectures. The identification performance of seven models was evaluated and compared with that of six physicians. RESULTS: All models exhibited a high accuracy, precision, and specificity in recognizing NS patients. The VGG19-BN model delivered the best overall performance, with an accuracy of 93.76%, precision of 91.40%, specificity of 98.73%, and F1 score of 78.34%. The VGG16-BN model achieved the highest AUC value of 0.9787, while all models based on VGG architectures were superior to the others on the whole. The highest scores of six physicians in terms of accuracy, precision, specificity, and the F1 score were 74.00%, 75.00%, 88.33%, and 61.76%, respectively. The performance of each model of facial recognition was superior to that of the best physician on all metrics. CONCLUSION: Models of computer-assisted facial recognition can improve the rate of diagnosis of NS. The models based on VGG19-BN and VGG16-BN can play an important role in diagnosing NS in clinical practice.
Asunto(s)
Síndrome de Noonan , Humanos , Síndrome de Noonan/diagnóstico , Niño , Femenino , Masculino , Preescolar , Redes Neurales de la Computación , Lactante , Adolescente , Reconocimiento Facial Automatizado/métodos , Diagnóstico por Computador/métodos , Sensibilidad y Especificidad , Estudios de Casos y ControlesRESUMEN
Batch Normalization's (BN) unique property of depending on other samples in a batch is known to cause problems in several tasks, including sequence modeling. Yet, BN-related issues are hardly studied for long video understanding, despite the ubiquitous use of BN in CNNs (Convolutional Neural Networks) for feature extraction. Especially in surgical workflow analysis, where the lack of pretrained feature extractors has led to complex, multi-stage training pipelines, limited awareness of BN issues may have hidden the benefits of training CNNs and temporal models end to end. In this paper, we analyze pitfalls of BN in video learning, including issues specific to online tasks such as a 'cheating' effect in anticipation. We observe that BN's properties create major obstacles for end-to-end learning. However, using BN-free backbones, even simple CNN-LSTMs beat the state of the art on three surgical workflow benchmarks by utilizing adequate end-to-end training strategies which maximize temporal context. We conclude that awareness of BN's pitfalls is crucial for effective end-to-end learning in surgical tasks. By reproducing results on natural-video datasets, we hope our insights will benefit other areas of video learning as well. Code is available at: https://gitlab.com/nct_tso_public/pitfalls_bn.
Asunto(s)
Redes Neurales de la Computación , Humanos , Flujo de TrabajoRESUMEN
Spiking neural networks (SNNs) have garnered significant attention due to their computational patterns resembling biological neural networks. However, when it comes to deep SNNs, how to focus on critical information effectively and achieve a balanced feature transformation both temporally and spatially becomes a critical challenge. To address these challenges, our research is centered around two aspects: structure and strategy. Structurally, we optimize the leaky integrate-and-fire (LIF) neuron to enable the leakage coefficient to be learnable, thus making it better suited for contemporary applications. Furthermore, the self-attention mechanism is introduced at the initial time step to ensure improved focus and processing. Strategically, we propose a new normalization method anchored on the learnable leakage coefficient (LLC) and introduce a local loss signal strategy to enhance the SNN's training efficiency and adaptability. The effectiveness and performance of our proposed methods are validated on the MNIST, FashionMNIST, and CIFAR-10 datasets. Experimental results show that our model presents a superior, high-accuracy performance in just eight time steps. In summary, our research provides fresh insights into the structure and strategy of SNNs, paving the way for their efficient and robust application in practical scenarios.
RESUMEN
In order to fully explore the neural oscillatory coupling characteristics of patients with mild cognitive impairment (MCI), this paper analyzed and compared the strength of the coupling characteristics for 28 MCI patients and 21 normal subjects under six different-frequency combinations. The results showed that the difference in the global phase synchronization index of cross-frequency coupling under δ-θ rhythm combination was statistically significant in the MCI group compared with the normal control group ( P = 0.025, d = 0.398). To further validate this coupling feature, this paper proposed an optimized convolutional neural network model that incorporated a time-frequency data enhancement module and batch normalization layers to prevent overfitting while enhancing the robustness of the model. Based on this optimized model, with the phase locking value matrix of δ-θ rhythm combination as the single input feature, the diagnostic accuracy of MCI patients was (95.49 ± 4.15)%, sensitivity and specificity were (93.71 ± 7.21)% and (97.50 ± 5.34)%, respectively. The results showed that the characteristics of the phase locking value matrix under the combination of δ-θ rhythms can adequately reflect the cognitive status of MCI patients, which is helpful to assist the diagnosis of MCI.
Asunto(s)
Disfunción Cognitiva , Electroencefalografía , Humanos , Electroencefalografía/métodos , Disfunción Cognitiva/diagnóstico , Redes Neurales de la Computación , Sensibilidad y EspecificidadRESUMEN
Convolutional neural networks (CNNs) have been extensively utilized in medical image processing to automatically extract meaningful features and classify various medical conditions, enabling faster and more accurate diagnoses. In this paper, LeNet, a classic CNN architecture, has been successfully applied to breast cancer data analysis. It demonstrates its ability to extract discriminative features and classify malignant and benign tumors with high accuracy, thereby supporting early detection and diagnosis of breast cancer. LeNet with corrected Rectified Linear Unit (ReLU), a modification of the traditional ReLU activation function, has been found to improve the performance of LeNet in breast cancer data analysis tasks via addressing the "dying ReLU" problem and enhancing the discriminative power of the extracted features. This has led to more accurate, reliable breast cancer detection and diagnosis and improved patient outcomes. Batch normalization improves the performance and training stability of small and shallow CNN architecture like LeNet. It helps to mitigate the effects of internal covariate shift, which refers to the change in the distribution of network activations during training. This classifier will lessen the overfitting problem and reduce the running time. The designed classifier is evaluated against the benchmarking deep learning models, proving that this has produced a higher recognition rate. The accuracy of the breast image recognition rate is 89.91%. This model will achieve better performance in segmentation, feature extraction, classification, and breast cancer tumor detection.
RESUMEN
Red, blue, white, pink, or black spots with irregular borders and small lesions on the skin are known as skin cancer that is categorized into two types: benign and malignant. Skin cancer can lead to death in advanced stages, however, early detection can increase the chances of survival of skin cancer patients. There exist several approaches developed by researchers to identify skin cancer at an early stage, however, they may fail to detect the tiniest tumours. Therefore, we propose a robust method for the diagnosis of skin cancer, namely SCDet, based on a convolutional neural network (CNN) having 32 layers for the detection of skin lesions. The images, having a size of 227 × 227, are fed to the image input layer, and then pair of convolution layers is utilized to withdraw the hidden patterns of the skin lesions for training. After that, batch normalization and ReLU layers are used. The performance of our proposed SCDet is computed using the evaluation matrices: precision 99.2%; recall 100%; sensitivity 100%; specificity 99.20%; and accuracy 99.6%. Moreover, the proposed technique is compared with the pre-trained models, i.e., VGG16, AlexNet, and SqueezeNet and it is observed that SCDet provides higher accuracy than these pre-trained models and identifies the tiniest skin tumours with maximum precision. Furthermore, our proposed model is faster than the pre-trained model as the depth of its architecture is not too high as compared to pre-trained models such as ResNet50. Additionally, our proposed model consumes fewer resources during training; therefore, it is better in terms of computational cost than the pre-trained models for the detection of skin lesions.
RESUMEN
The field of neuroimaging has increasingly sought to develop artificial intelligence-based models for neurological and neuropsychiatric disorder automated diagnosis and clinical decision support. However, if these models are to be implemented in a clinical setting, transparency will be vital. Two aspects of transparency are (1) confidence estimation and (2) explainability. Confidence estimation approaches indicate confidence in individual predictions. Explainability methods give insight into the importance of features to model predictions. In this study, we integrate confidence estimation and explainability approaches for the first time. We demonstrate their viability for schizophrenia diagnosis using resting state functional magnetic resonance imaging (rs-fMRI) dynamic functional network connectivity (dFNC) data. We compare two confidence estimation approaches: Monte Carlo dropout (MCD) and MC batch normalization (MCBN). We combine them with two gradient-based explainability approaches, saliency and layer-wise relevance propagation (LRP), and examine their effects upon explanations. We find that MCD often adversely affects model gradients, making it ill-suited for integration with gradient-based explainability methods. In contrast, MCBN does not affect model gradients. Additionally, we find many participant-level differences between regular explanations and the distributions of explanations for combined explainability and confidence estimation approaches. This suggests that a similar confidence estimation approach used in a clinical context with explanations only output for the regular model would likely not yield adequate explanations. We hope that our findings will provide a starting point for the integration of the two fields, provide useful guidance for future studies, and accelerate the development of transparent neuroimaging clinical decision support systems.
RESUMEN
Introduction: Crop pests have a great impact on the quality and yield of crops. The use of deep learning for the identification of crop pests is important for crop precise management. Methods: To address the lack of data set and poor classification accuracy in current pest research, a large-scale pest data set named HQIP102 is built and the pest identification model named MADN is proposed. There are some problems with the IP102 large crop pest dataset, such as some pest categories are wrong and pest subjects are missing from the images. In this study, the IP102 data set was carefully filtered to obtain the HQIP102 data set, which contains 47,393 images of 102 pest classes on eight crops. The MADN model improves the representation capability of DenseNet in three aspects. Firstly, the Selective Kernel unit is introduced into the DenseNet model, which can adaptively adjust the size of the receptive field according to the input and capture target objects of different sizes more effectively. Secondly, in order to make the features obey a stable distribution, the Representative Batch Normalization module is used in the DenseNet model. In addition, adaptive selection of whether to activate neurons can improve the performance of the network, for which the ACON activation function is used in the DenseNet model. Finally, the MADN model is constituted by ensemble learning. Results: Experimental results show that MADN achieved an accuracy and F1Score of 75.28% and 65.46% on the HQIP102 data set, an improvement of 5.17 percentage points and 5.20 percentage points compared to the pre-improvement DenseNet-121. Compared with ResNet-101, the accuracy and F1Score of MADN model improved by 10.48 percentage points and 10.56 percentage points, while the parameters size decreased by 35.37%. Deploying models to cloud servers with mobile application provides help in securing crop yield and quality.
RESUMEN
Background and Objective: Cardiovascular disease is a high-fatality health issue. Accurate measurement of cardiovascular function depends on precise segmentation of physiological structure and accurate evaluation of functional parameters. Structural segmentation of heart images and calculation of the volume of different ventricular activity cycles form the basis for quantitative analysis of physiological function and can provide the necessary support for clinical physiological diagnosis, as well as the analysis of various cardiac diseases. Therefore, it is important to develop an efficient heart segmentation algorithm. Methods: A total of 275 nuclear magnetic resonance imaging (MRI) heart scans were collected, analyzed, and preprocessed from Huaqiao University Affiliated Strait Hospital, and the data were used in our improved deep learning model, which was designed based on the U-net network. The training set included 80% of the images, and the remaining 20% was the test set. Based on five time phases from end-diastole (ED) to end-systole (ES), the segmentation findings showed that it is possible to achieve improved segmentation accuracy and computational complexity by segmenting the left ventricle (LV), right ventricle (RV), and myocardium (myo). Results: We improved the Dice index of the LV to 0.965 and 0.921, and the Hausdorff index decreased to 5.4 and 6.9 in the ED and ES phases, respectively; RV Dice increased to 0.938 and 0.860, and the Hausdorff index decreased to 11.7 and 12.6 in the ED and ES, respectively; myo Dice increased to 0.889 and 0.901, and the Hausdorff index decreased to 8.3 and 9.2 in the ED and ES, respectively. Conclusion: The model obtained in the final experiment provided more accurate segmentation of the left and right ventricles, as well as the myocardium, from cardiac MRI. The data from this model facilitate the prediction of cardiovascular disease in real-time, thereby providing potential clinical utility.
RESUMEN
In the biometric field, vein identification is a vital process that is constrained by the invisibility of veins as well as other unique features. Moreover, users generally do not wish to have their personal information uploaded to the cloud, so edge computing has become popular for the sake of protecting user privacy. In this paper, we propose a low-complexity and lightweight convolutional neural network (CNN) and we design intellectual property (IP) for shortening the inference time in finger vein recognition. This neural network system can operate independently in client mode. After fetching the user's finger vein image via a near-infrared (NIR) camera mounted on an embedded system, vein features can be efficiently extracted by vein curving algorithms and user identification can be completed quickly. Better image quality and higher recognition accuracy can be obtained by combining several preprocessing techniques and the modified CNN. Experimental data were collected by the finger vein image capture equipment developed in our laboratory based on the specifications of similar products currently on the market. Extensive experiments demonstrated the practicality and robustness of the proposed finger vein identification system.
Asunto(s)
Algoritmos , Redes Neurales de la Computación , Humanos , Biometría , Extremidades , LaboratoriosRESUMEN
Unsupervised domain adaptation (UDA) has been a vital protocol for migrating information learned from a labeled source domain to facilitate the implementation in an unlabeled heterogeneous target domain. Although UDA is typically jointly trained on data from both domains, accessing the labeled source domain data is often restricted, due to concerns over patient data privacy or intellectual property. To sidestep this, we propose "off-the-shelf (OS)" UDA (OSUDA), aimed at image segmentation, by adapting an OS segmentor trained in a source domain to a target domain, in the absence of source domain data in adaptation. Toward this goal, we aim to develop a novel batch-wise normalization (BN) statistics adaptation framework. In particular, we gradually adapt the domain-specific low-order BN statistics, e.g., mean and variance, through an exponential momentum decay strategy, while explicitly enforcing the consistency of the domain shareable high-order BN statistics, e.g., scaling and shifting factors, via our optimization objective. We also adaptively quantify the channel-wise transferability to gauge the importance of each channel, via both low-order statistics divergence and a scaling factor. Furthermore, we incorporate unsupervised self-entropy minimization into our framework to boost performance alongside a novel queued, memory-consistent self-training strategy to utilize the reliable pseudo label for stable and efficient unsupervised adaptation. We evaluated our OSUDA-based framework on both cross-modality and cross-subtype brain tumor segmentation and cardiac MR to CT segmentation tasks. Our experimental results showed that our memory consistent OSUDA performs better than existing source-relaxed UDA methods and yields similar performance to UDA methods with source data.
Asunto(s)
Neoplasias Encefálicas , Aprendizaje , Humanos , Neoplasias Encefálicas/diagnóstico por imagen , Entropía , Corazón , Movimiento (Física)RESUMEN
Frequency is essential in signal transmission, especially in convolutional neural networks. It is vital to maintain the signal frequency in the neural network to maintain the performance of a convolutional neural network. Due to destructive signal transmission in convolutional neural network, signal frequency downconversion in channels results into incomplete spatial information. In communication theory, the number of Fourier series coefficients determines the integrity of the information transmitted in channels. Consequently, the number of Fourier series coefficients of the signals can be replenished to reduce the information transmission loss. To achieve this, the ArsenicNetPlus neural network was proposed for signal transmission modulation in detecting cassava diseases. First, multiattention was used to maintain the long-term dependency of the features of cassava diseases. Afterward, depthwise convolution was implemented to remove aliasing signals and downconvert before the sampling operation. Instance batch normalization algorithm was utilized to keep features in an appropriate form in the convolutional neural network channels. Finally, the ArsenicPlus block was implemented to generate pseudo high-frequency in the residual structure. The proposed method was tested on the Cassava Datasets and compared with the V2-ResNet-101, EfficientNet-B5, RepVGG-B3g4 and AlexNet. The results showed that the proposed method performed [Formula: see text] in terms of accuracy, 1.2440 in terms of loss, and [Formula: see text] in terms of the F1-score, outperforming the comparison algorithms.
RESUMEN
Earthquakes threaten people, homes, and infrastructure. Early warning systems provide prior warning of oncoming significant shaking to decrease seismic risk by providing location, magnitude, and depth information of the event. Their usefulness depends on how soon a strong shake begins after the warning. In this article, the authors implement a deep learning model for predicting earthquakes. This model is based on a graph convolutional neural network with batch normalization and attention mechanism techniques that can successfully predict the depth and magnitude of an earthquake event at any number of seismic stations in any number of locations. After preprocessing the waveform data, CNN extracts the feature map. Attention mechanism is used to focus on important features. The batch normalization technique takes place in batches for stable and faster training of the model by adding an extra layer. GNN with extracted features and event location information predicts the event information accurately. We test the proposed model on two datasets from Japan and Alaska, which have different seismic dynamics. The proposed model achieves 2.8 and 4.0 RMSE values in Alaska and Japan for magnitude prediction, and 2.87 and 2.66 RMSE values for depth prediction. Low RMSE values show that the proposed model significantly outperforms the three baseline models on both datasets to provide an accurate estimation of the depth and magnitude of small, medium, and large-magnitude events.
Asunto(s)
Terremotos , Humanos , Japón , Redes Neurales de la ComputaciónRESUMEN
The ball screw is the core component of the CNC machine tool feed system, and its health plays an important role in the feed system and even in the entire CNC machine tool. This paper studies the fault diagnosis and health assessment of ball screws. Aiming at the problem that the ball screw signal is weak and susceptible to interference, using a wavelet convolution structure to improve the network can improve the mining ability of signal time domain and frequency domain features; aiming at the challenge of ball screw sensor installation position limitation, a transfer learning method is proposed, which adopts the domain adaptation method as jointly distributed adaptation (JDA), and realizes the transfer diagnosis across measurement positions by extracting the diagnosis knowledge of different positions of the ball screw. In this paper, the adaptive batch normalization algorithm (AdaBN) is introduced to enhance the proposed model so as to improve the accuracy of migration diagnosis. Experiments were carried out using a self-made lead screw fatigue test bench. Through experimental verification, the method proposed in this paper can extract effective fault diagnosis knowledge. By collecting data under different working conditions at the bearing seat of the ball screw, the fault diagnosis knowledge is extracted and used to identify and diagnose the position fault of the nut seat. In this paper, some background noise is added to the collected data to test the robustness of the proposed network model.
Asunto(s)
Algoritmos , Ruido , Aprendizaje AutomáticoRESUMEN
BACKGROUND: Circular RNAs (CircRNAs) play critical roles in gene expression regulation and disease development. Understanding the regulation mechanism of CircRNAs formation can help reveal the role of CircRNAs in various biological processes mentioned above. Back-splicing is important for CircRNAs formation. Back-splicing sites prediction helps uncover the mysteries of CircRNAs formation. Several methods were proposed for back-splicing sites prediction or circRNA-realted prediction tasks. Model performance was constrained by poor feature learning and using ability. RESULTS: In this study, CircCNN was proposed to predict pre-mRNA back-splicing sites. Convolution neural network and batch normalization are the main parts of CircCNN. Experimental results on three datasets show that CircCNN outperforms other baseline models. Moreover, PPM (Position Probability Matrix) features extract by CircCNN were converted as motifs. Further analysis reveals that some of motifs found by CircCNN match known motifs involved in gene expression regulation, the distribution of motif and special short sequence is important for pre-mRNA back-splicing. CONCLUSIONS: In general, the findings in this study provide a new direction for exploring CircRNA-related gene expression regulatory mechanism and identifying potential targets for complex malignant diseases. The datasets and source code of this study are freely available at: https://github.com/szhh521/CircCNN .
Asunto(s)
Precursores del ARN , ARN Circular , Regulación de la Expresión Génica , Redes Neurales de la Computación , Precursores del ARN/metabolismo , Empalme del ARNRESUMEN
Various lengths of time window have been used in feature extraction for electroencephalogram (EEG) signal processing in previous studies. However, the effect of time window length on feature extraction for the downstream tasks such as emotion recognition has not been well examined. To this end, we investigate the effect of different time window (TW) lengths on human emotion recognition to find the optimal TW length for extracting electroencephalogram (EEG) emotion signals. Both power spectral density (PSD) features and differential entropy (DE) features are used to evaluate the effectiveness of different TW lengths based on the SJTU emotion EEG dataset (SEED). Different lengths of TW are then processed with an EEG feature-processing approach, namely experiment-level batch normalization (ELBN). The processed features are used to perform emotion recognition tasks in the six classifiers, the results of which are then compared with the results without ELBN. The recognition accuracies indicate that a 2-s TW length has the best performance on emotion recognition and is the most suitable to be used in EEG feature extraction for emotion recognition. The deployment of ELBN in the 2-s TW can further improve the emotion recognition performances by 21.63% and 5.04% when using an SVM based on PSD and DE features, respectively. These results provide a solid reference for the selection of TW length in analyzing EEG signals for applications in intelligent systems.
Asunto(s)
Electroencefalografía , Procesamiento de Señales Asistido por Computador , Electroencefalografía/métodos , Emociones , Humanos , Reconocimiento en PsicologíaRESUMEN
Recently, with the rapid development of artificial intelligence, image generation based on deep learning has advanced significantly. Image generation based on Generative Adversarial Networks (GANs) is a promising study. However, because convolutions are limited by spatial-agnostic and channel-specific, features extracted by conventional GANs based on convolution are constrained. Therefore, GANs cannot capture in-depth details per image. Moreover, straightforwardly stacking of convolutions causes too many parameters and layers in GANs, yielding a high overfitting risk. To overcome the abovementioned limitations, in this study, we propose a GANs called GIU-GANs (where Global Information Utilization: GIU). GIU-GANs leverages a new module called the GIU module, which integrates the squeeze-and-excitation module and involution to focus on global information via the channel attention mechanism, enhancing the generated image quality. Moreover, Batch Normalization (BN) inevitably ignores the representation differences among noise sampled by the generator and thus degrades the generated image quality. Thus, we introduce the representative BN to the GANs' architecture. The CIFAR-10 and CelebA datasets are employed to demonstrate the effectiveness of the proposed model. Numerous experiments indicate that the proposed model achieves state-of-the-art performance.
Asunto(s)
Procesamiento de Imagen Asistido por Computador , Redes Neurales de la Computación , Inteligencia Artificial , Procesamiento de Imagen Asistido por Computador/métodosRESUMEN
Current designs of advanced driving assistance systems (ADAS) mainly developed uniform collision warning algorithms, which ignore the heterogeneity of driving behaviors, thus lead to low drivers' trust in. To address this issue, developing personalized driving assistance algorithms is a promising approach. However, current personalization systems were mainly implemented through manually adjusting warning trigger thresholds, which would be less feasible for overall drivers as certain domain expertise is required to set personal thresholds accurately. Other personalization techniques exploited individual drivers' data to build personalized models. Such approach could learn personal behavior but requires impractical large-scale individual data collections. To fill up the gaps, self-adaptive algorithms for personalized forward collision warning (FCW) based on federated learning were proposed in this study. A baseline model was developed by long short-term memory (LSTM) for FCW. Federated learning framework was then introduced to collect knowledge from multiple drivers with privacy preserving. Specifically, a general cloud server model was trained by collecting updated parameters from individual vehicle server models rather than collecting raw data. Besides, a driver-specific batch normalization (BN) layer was added into each vehicle server model to address the heterogeneity of driving behaviors. Experiments show empirically that the proposed federated-based personalized models with the BN layer showed to have the best performance. The average modeling accuracy has reached 84.88% and the performance is comparable to conventional total data collection training approach, where the additional BN layer could increase the accuracy by 3.48%. Finally, applications of the proposed framework and its further investigations have been discussed.
Asunto(s)
Accidentes de Tránsito , Algoritmos , Conducción de Automóvil , Accidentes de Tránsito/prevención & control , Humanos , Aprendizaje Automático , Equipos de Seguridad , Reproducibilidad de los ResultadosRESUMEN
With the improvement of industrial requirements for the quality of cold rolled strips, flatness has become one of the most important indicators for measuring the quality of cold rolled strips. In this paper, the strip production data of a 1250 mm tandem cold mill in a steel plant is modeled by an improved deep neural network (the improved DNN) to improve the accuracy of strip shape prediction. Firstly, the type of activation function is analyzed, and the monotonicity of the activation function is deemed independent of the convexity of the loss function in the deep network. Regardless of whether the activation function is monotonic, the loss function is not strictly convex. Secondly, the non-convex optimization of the loss functionextended from the deep linear network to the deep nonlinear network, is discussed, and the critical point of the deep nonlinear network is identified as the global minimum point. Finally, an improved Swish activation function based on batch normalization is proposed, and its performance is evaluated on the MNIST dataset. The experimental results show that the loss of an improved Swish function is lower than that of other activation functions. The prediction accuracy of a deep neural network (DNN) with an improved Swish function is 0.38% more than that of a deep neural network (DNN) with a regular Swish function. For the DNN with the improved Swish function, the mean square error of the prediction for the flatness of cold rolled strip is reduced to 65% of the regular DNN. The accuracy of the improved DNN is up to and higher than the industrial requirements. The shape prediction of the improved DNN will assist and guide the industrial production process, reducing the scrap yield and industrial cost.