Deep Neural Model Research Articles

Microscopic identification of foodborne bacterial pathogens based on deep learning method

Accurate and rapid detection of foodborne bacterial pathogens is critical for food quality control. Nowadays, tracking morphological bacterial properties using microscope is still a priority at the grass-roots food supervision department due to its simplicity and low cost. However, the method requires highly professional personnel and there are certain misjudgments in the process of analysis. Automatically recognizing foodborne pathogen using deep learning algorithm to replace manual microscopy will not only reduce expert cost, artificial misjudgment, and operation time in detection, but also provide more objective and accurate identification. Here, we firstly constructed a high-quality and large-scale dataset of foodborne pathogenic bacteria, allowing the deep learning-based model to be efficiently trained and achieve accurate identification. The deep convolutional neural network-based model is capable of identifying six common foodborne pathogens, including Escherichia coli (O157:H7), Vibrio parahaemolyticus, Staphylococcus aureus, Bacillus cereus, Salmonella typhi, and Streptococcus hemolyticus, with accuracy rates of 90%–100%. This method can assist or replace the manual microscopic inspection step in traditional detection methods, and is promising to break through the traditional approach that heavily relies on manual judgment, greatly reduce the cost of experts and human errors, and provide rapid, accurate, and powerful discriminatory results in large quantities for detection.

Techno-economic ionic liquid-based capturing, electrochemical reduction, and hydrogenation of carbon dioxide in the simultaneous production of formic acid and biomethane

CO2 utilization is vital for mitigating climate change by converting CO2 into valuable products, promoting environmental protection and resource efficiency. Novel pathways for CO2 utilization to produce formic acid are proposed namely solute phase electroreduction, gas phase electroreduction, and hydrogenation, are investigated. Employing multi-objective optimization with a deep neural network surrogate model, this study identifies optimal process conditions balancing capital and operational expenditures. The result shows that CO2 hydrogenation ($868 ton−1) exhibits the lowest production cost followed by gas-phase electroreduction ($986 ton−1) and solute-phase electroreduction ($2103 ton−1). The result also shows that without any intervention at all only hydrogenation can generate profit. Furthermore, an in-depth analysis of CO2 emissions indicates that gas phase electroreduction results in the lowest CO2 emissions (0.7 kg CO2 kgHCOOH−1) among the examined pathways. Insights from our research suggest a minimum current density of 417.3 mA cm−2 is recommended to achieve at least parity with hydrogenation in terms of production cost. To push the commercialization of gas-phase electroreduction, besides current density improvement, electricity cost reduction, and carbon trading mechanism is proposed to reduce the production cost.

Enhancing wrist abnormality detection with YOLO: Analysis of state-of-the-art single-stage detection models

Diagnosing and treating abnormalities in the wrist, specifically distal radius, and ulna fractures, is a crucial concern among children, adolescents, and young adults, with a higher incidence rate during puberty. However, the scarcity of radiologists and the lack of specialized training among medical professionals pose a significant risk to patient care. This problem is further exacerbated by the rising number of imaging studies and limited access to specialist reporting in certain regions. This highlights the need for innovative solutions to improve the diagnosis and treatment of wrist abnormalities. Automated wrist fracture detection using object detection has shown potential, but current studies mainly use two-stage detection methods with limited evidence for single-stage effectiveness. This study employs state-of-the-art single-stage deep neural network-based detection models YOLOv5, YOLOv6, YOLOv7, and YOLOv8 to detect wrist abnormalities. Through extensive experimentation, we found that these YOLO models outperform the commonly used two-stage detection algorithm, Faster R-CNN, in fracture detection. Additionally, compound-scaled variants of each YOLO model were compared, with YOLOv8 m demonstrating a highest fracture detection sensitivity of 0.92 and mean average precision (mAP) of 0.95. On the other hand, YOLOv6 m achieved the highest sensitivity across all classes at 0.83. Meanwhile, YOLOv8x recorded the highest mAP of 0.77 for all classes on the GRAZPEDWRI-DX pediatric wrist dataset, highlighting the potential of single-stage models for enhancing pediatric wrist imaging.

Recognition of eye diseases based on deep neural networks for transfer learning and improved D-S evidence theory

BackgroundHuman vision has inspired significant advancements in computer vision, yet the human eye is prone to various silent eye diseases. With the advent of deep learning, computer vision for detecting human eye diseases has gained prominence, but most studies have focused only on a limited number of eye diseases.ResultsOur model demonstrated a reduction in inherent bias and enhanced robustness. The fused network achieved an Accuracy of 0.9237, Kappa of 0.878, F1 Score of 0.914 (95% CI [0.875–0.954]), Precision of 0.945 (95% CI [0.928–0.963]), Recall of 0.89 (95% CI [0.821–0.958]), and an AUC value of ROC at 0.987. These metrics are notably higher than those of comparable studies.ConclusionsOur deep neural network-based model exhibited improvements in eye disease recognition metrics over models from peer research, highlighting its potential application in this field.MethodsIn deep learning-based eye recognition, to improve the learning efficiency of the model, we train and fine-tune the network by transfer learning. In order to eliminate the decision bias of the models and improve the credibility of the decisions, we propose a model decision fusion method based on the D-S theory. However, D-S theory is an incomplete and conflicting theory, we improve and eliminate the existed paradoxes, propose the improved D-S evidence theory(ID-SET), and apply it to the decision fusion of eye disease recognition models.

Explaining Bounding Boxes in Deep Object Detectors Using Post Hoc Methods for Autonomous Driving Systems.

Deep learning has rapidly increased in popularity, leading to the development of perception solutions for autonomous driving. The latter field leverages techniques developed for computer vision in other domains for accomplishing perception tasks such as object detection. However, the black-box nature of deep neural models and the complexity of the autonomous driving context motivates the study of explainability in these models that perform perception tasks. Moreover, this work explores explainable AI techniques for the object detection task in the context of autonomous driving. An extensive and detailed comparison is carried out between gradient-based and perturbation-based methods (e.g., D-RISE). Moreover, several experimental setups are used with different backbone architectures and different datasets to observe the influence of these aspects in the explanations. All the techniques explored consist of saliency methods, making their interpretation and evaluation primarily visual. Nevertheless, numerical assessment methods are also used. Overall, D-RISE and guided backpropagation obtain more localized explanations. However, D-RISE highlights more meaningful regions, providing more human-understandable explanations. To the best of our knowledge, this is the first approach to obtaining explanations focusing on the regression of the bounding box coordinates.

Safe path planning and adjustable zonotope-tube model predictive tracking control for autonomous vehicle

The path planning and tracking control problem of autonomous vehicle with model matched uncertainty and external disturbance is studied in this paper. A safe lane change zone is designed by considering lane change time, distance, and driving speed. Meanwhile, the candidate paths are generated using cubic quasi-uniform B-spline curve, and the optimal path is selected based on feasibility, comfort, and efficiency criteria. The proposed planning method ensures an efficient and comfortable lane change path without collision with surrounding vehicles. Model uncertainty and external disturbance may affect lane change tracking control of autonomous vehicle. Hence, a fault-tolerant path tracking controller is implemented by using deep neural networks and adjustable zonotope-tube model predictive control. A deep neural network-based controller is proposed to compensate for model matched uncertainty, thereby enhancing the fault tolerance of the control system. Then, an adjustable zonotope-tube model predictive controller is designed to enclose the actual trajectory within a zonotope-tube with the nominal state as the center and the disturbance set as the radius by using the feedback control to achieve the asymptotic stabilization of the closed-loop system. The effectiveness of this path planning and tracking control method is finally verified by numerical simulation.

Trajectory tracking control of autonomous vehicles based on Lagrangian neural network dynamics model

The autonomous vehicles make decisions and plans based on the environmental perception and generate the target command of the control layer. The vehicle dynamics model is an important factor that affects the vehicle control. The dynamic mechanism model has strong interpretability and good stability. However, in extreme conditions, the model accuracy is reduced due to the tire entering the nonlinear region. The data-driven dynamic model achieves high modeling accuracy. However, due to the lack of physical constraints and rationality in the data-driven models, the interpretability and stability of the control is reduced, which in turn increases the unpredictable risk in the driving process. This paper innovatively proposes a deep Lagrangian neural network dynamics model (DeLaN) for autonomous vehicles based on the Lagrangian mechanics and uses a neural network to encode the differential equations. This not only retains the interpretability of the physical model but also makes full use of the learning ability and fitting ability of the neural network to effectively capture the complex dynamic characteristics of the vehicle. To improve the robustness of the control system, this work uses DeLaN as feed-forward control and preview error feedback control to form a closed loop of trajectory tracking control for autonomous vehicles. The experimental results show that the trajectory tracking error of the proposed DeLaN is significantly reduced, the yaw stability and comfort are significantly improved, good longitudinal and lateral cooperative control performance is achieved, and the physical rationality of the neural network is also improved. Therefore, the proposed DeLaN has important engineering application value.

Quantitative Analysis of the Impact of Different Delivery Modes on Cervical Elasticity Based on Real-time Shear Wave Imaging Technology and Artificial Intelligence.

Real-time shear wave elastography (SWE) has emerged as a useful imaging modality for detecting the lesion location of various diseases, including cervical diseases. In this paper, the SWE was used to quantitatively determine the tissue hardness of the internal and external orifice of the cervix (IOC & EOC) and to relatively objectively analyze the impact of different production methods on the hardness of the cervical tissue. A total of 48 patients were selected, and they were divided into three groups according to different production methods: control group (16 cases), cesarean section group (16 cases), and spontaneous delivery group (16 cases). Artificial intelligence has also been incorporated into this work. A deep flexible neural tree model and a new set of FNT models were proposed to assist in classifying cervical physical data in different states. The physical data was extracted as the features, and the different states were considered as category labels. There was no statistically significant difference in the elasticity of the IOC and the EOC between the groups. However, the difference in the elasticity of the IOC and the EOC within each group was statistically significant. The classification results corresponded with the results of the statistical analysis. The hardness of the EOC is generally lower than that of the IOC, and there was no significant difference in hardness between the IOC and the EOC in the three groups. There is no significant difference in the cervical elasticity hardness between different delivery modes.

Kernelized Deep Learning for Matrix Factorization Recommendation System Using Explicit and Implicit Information.

In the current matrix factorization recommendation approaches, the item and the user latent factor vectors are with the same dimension. Thus, the linear dot product is used as the interactive function between the user and the item to predict the ratings. However, the relationship between real users and items is not entirely linear and the existing recommendation model of matrix factorization faces the challenge of data sparsity. To this end, we propose a kernelized deep neural network recommendation model in this article. First, we encode the explicit user-item rating matrix in the form of column vectors and project them to higher dimensions to facilitate the simulation of nonlinear user-item interaction for enhancing the connection between users and items. Second, the algorithm of association rules is used to mine the implicit relation between users and items, rather than simple feature extraction of users or items, for improving the recommendation performance when the datasets are sparse. Third, through the autoencoder and kernelized network processing, the implicit data are connected with the explicit data by the multilayer perceptron network for iterative training instead of doing simple linear weighted summation. Finally, the predicted rating is output through the hidden layer. Extensive experiments were conducted on four public datasets in comparison with several existing well-known methods. The experimental results indicated that our proposed method has obtained improved performance in data sparsity and prediction accuracy.

Optimization and Integration of Student Work Management Mechanism in Colleges and Universities Based on Deep Neural Networks

Abstract Under the new situation, student management work in colleges and universities must keep pace with the development of information technology, take the initiative to comply with the laws of the new era, innovate work concepts, revolutionize work methods, and make full use of the new era technology to improve the level of management services. After integrating and optimizing the management mechanism of student work in colleges and universities, the study establishes a deep neural network-based student management work effect assessment model in colleges and universities using cluster analysis and convolutional neural network and takes 20 colleges and universities in city A as examples to conduct empirical analysis. The results show that the fit of the CNN neural network model is relatively good, and the total R value is 0.99634, R 2 = 0.98324, R 2 is greater than 0.75. The student work management mechanism has been optimized and integrated by 13 colleges and universities, and their scores are all above 80 points. Among them, there are 5 colleges and universities whose evaluation results exceed 90 points. The construction of their student management cadres, the construction of student self-government capacity, the construction of school spirit and academic style, the work of campus security and stability, and the campus cultural activities have achieved better optimization results.

A Hybrid Deep Implicit Neural Model for Sentiment Analysis via Transfer Learning

A Hybrid Deep Implicit Neural Model for Sentiment Analysis via Transfer Learning

Learning Virtual View Selection for 3D Scene Semantic Segmentation.

2D-3D joint learning is essential and effective for fundamental 3D vision tasks, such as 3D semantic segmentation, due to the complementary information these two visual modalities contain. Most current 3D scene semantic segmentation methods process 2D images "as they are", i.e., only real captured 2D images are used. However, such captured 2D images may be redundant, with abundant occlusion and/or limited field of view (FoV), leading to poor performance for the current methods involving 2D inputs. In this paper, we propose a general learning framework for joint 2D-3D scene understanding by selecting informative virtual 2D views of the underlying 3D scene. We then feed both the 3D geometry and the generated virtual 2D views into any joint 2D-3D-input or pure 3D-input based deep neural models for improving 3D scene understanding. Specifically, we generate virtual 2D views based on an information score map learned from the current 3D scene semantic segmentation results. To achieve this, we formalize the learning of the information score map as a deep reinforcement learning process, which rewards good predictions using a deep neural network. To obtain a compact set of virtual 2D views that jointly cover informative surfaces of the 3D scene as much as possible, we further propose an efficient greedy virtual view coverage strategy in the normal-sensitive 6D space, including 3-dimensional point coordinates and 3-dimensional normal. We have validated our proposed framework for various joint 2D-3D-input or pure 3D-input based deep neural models on two real-world 3D scene datasets, i.e., ScanNet v2 and S3DIS, and the results demonstrate that our method obtains a consistent gain over baseline models and achieves new top accuracy for joint 2D and 3D scene semantic segmentation. Code is available at https://github.com/smy-THU/VirtualViewSelection.

Detection of cardiovascular disease using explainable artificial intelligence and gut microbiota data

Purpose:Gut microbiota are defined as the microbial population of the intestines. They include various types of bacteria which can influence and predict the existence or onset of some specific diseases. Therefore, it is a common practice in medicine to analyze the gut microbiota for diagnostic purposes by analyzing certain measurable biochemical features associated with the disease under investigation. However, the evaluation of all the data collected from the gut microbiota is a labor-intensive process. Artificial Intelligence (AI) may be a helpful tool to identify the hidden patterns in gut microbiota for the detection of disease and other classification problems. Methods:In this study, we propose a deep neural model based on a one-dimensional convolutional neural network (1D-CNN) to detect cardiovascular disease using bacterial taxonomy and OTU (Operational Taxonomic Unit) table data. The developed AI method is compared to classical machine learning algorithms, regression, boosting algorithms and a deep model, Tabular Network (TabNet), developed for tabular data and obtained outperforming classification results. Results:According to AUC (Area Under Curve) values, boosting and regression methods outperformed the classical machine learning methods. However, the highest value of 97.09 AUC was obtained with the developed 1D-CNN model by using bacterial taxonomy data even with less then expected number of samples. Using explainable AI, nine bacteria were identified which the models find important for classification. Conclusion:The proposed method is robust and well adapted to taxonomy data in tabular form. It can be easily adapted to detect other diseases by using taxonomy data. The study also investigated the effect on barcode sequence for the classification, but the result showed that barcode sequences do not contribute to the bacterial taxonomy data for the estimation of CVD disease.

Research on fault diagnosis method of linear vibration screen based on fused RP-improved CNN

A fault diagnosis method based on fused RP (recurrence plot) and improved CNN (convolutional neural networks) is proposed for the traditional fault diagnosis method of linear vibrating screen manually designed and optimized features with feature quality uncertainty. A deep convolutional neural network combination model (MDCNN) with high-level feature fusion is designed. The collected multi-source vibration signals were converted into black and white recurrence plot, and fusion into three-channel RGB recurrence plot. Fault diagnosis was classified by the combination model of deep convolution neural network with high level feature fusion. The results show that the recurrence plot of homologous multi-sensor signals contains more information about the characteristics of the vibrating screen than the original colorful signal map or the Gram Point Field, which contributes to the feature learning and classification of MDCNN, with an average recognition accuracy of 97.59% under strong vibrating sift-5db noise. Compared with manifold classical neural network models, the MDCNN model is proved to be effective, stable and small space occupying.

Radar Emitter Signal Intra-Pulse Modulation Open Set Recognition Based on Deep Neural Network

Radar emitter signal intra-pulse modulation recognition is important for modern electronic reconnaissance systems to analyze target radar systems. In the actual environment, the intra-pulse modulations of the sampled radar emitter signals contain not only the known types in the library but also the unknown types. Therefore, the existing recognition methods, which are based on a closed set, cannot recognize the unknown samples. In order to solve this problem, in this paper, we proposed a method for radar emitter signal intra-pulse modulation open set recognition. The proposed method could classify the known modulations and identify the unknown modulation by using an original deep neural network-based recognition model trained on a closed set, estimating the signal-to-noise ratio, and calculating the reconstruction loss by an encoder–decoder model. For a given sample, the original deep neural network-based recognition model will label it as a certain known class temporarily. By estimating the SNR of the sample and calculating the reconstruction loss by inputting the sample to the corresponding encoder–decoder model related to the temporary predicted known class, whether the sample belongs to the predicted temporary known class or the unknown class will be confirmed. Experiments were conducted with five different openness conditions. The experimental results indicate that the proposed method has good performance on radar emitter signal intra-pulse modulation open set recognition.

The Good, the Bad, and the Missing: Neural Code Generation for Machine Learning Tasks

Machine learning (ML) has been increasingly used in a variety of domains, while solving ML programming tasks poses unique challenges due to the fundamental difference in the nature and the construct of general programming tasks, especially for developers who do not have ML backgrounds. Automatic code generation that produces a code snippet from a natural language description can be a promising technique to accelerate ML programming tasks. In recent years, although many deep learning-based neural code generation models have been proposed with high accuracy, the fact that most of them are mainly evaluated on general programming tasks calls into question their effectiveness and usefulness in ML programming tasks. In this article, we set out to investigate the effectiveness of existing neural code generation models on ML programming tasks. For our analysis, we select six state-of-the-art neural code generation models and evaluate their performance on four widely used ML libraries, with newly created 83K pairs of natural-language described ML programming tasks. Our empirical study reveals some good, bad, and missing aspects of neural code generation models on ML tasks, with a few major ones listed below. ( Good ) Neural code generation models perform significantly better on ML tasks than on non-ML tasks with an average difference of 10.6 points in BLEU-4 scores. ( Bad ) More than 80% of the generated code is semantically incorrect. ( Bad ) Code generation models do not have significance in improving developers’ completion time. ( Good ) The generated code can help developers write correct code by providing developers with clues for using correct APIs. ( Missing ) The observation from our user study reveals the missing aspects of code generation for ML tasks, e.g., decomposing code generation for divide-and-conquer into API sequence identification and API usage generation.

A Deep Neural Network-Based Assistive Decision Method for Financial Risk Prediction in Carbon Trading Market

The price of carbon emission rights in the market fluctuates greatly due to various factors from economy, finance, and climate. For an enterprise that needs to conduct carbon trading activities, it is extremely necessary to fully grasp the price of carbon trading in the future period. Thus in this paper, a deep neural network-based assistive decision model for financial risk prediction in carbon trading market is proposed for this purpose. Specifically, the dynamic risk spillover effects of domestic and international carbon trading markets are studied, and a frontier time-varying model is utilized to measure the risk spillover effects. Then, the deep neural network is used for quantitative research and to construct an intelligent decision scheme that outputs financial risk prediction results. Four perspectives, energy price, climate environment, carbon market price, and macroeconomics, are selected as the input to analyze the influence factors of carbon emission rights price. Finally, several linear regression models are adopted as the baseline methods for comparison, and experimental results show that the proposed method can achieve better prediction performance compared with baseline methods.

Second-order Confidence Network for Early Classification of Time Series

Time series data are ubiquitous in a variety of disciplines. Early classification of time series, which aims to predict the class label of a time series as early and accurately as possible, is a significant but challenging task in many time-sensitive applications. Existing approaches mainly utilize heuristic stopping rules to capture stopping signals from the prediction results of time series classifiers. However, heuristic stopping rules can only capture obvious stopping signals, which makes these approaches give either correct but late predictions or early but incorrect predictions. To tackle the problem, we propose a novel second-order confidence network for early classification of time series, which can automatically learn to capture implicit stopping signals in early time series in a unified framework. The proposed model leverages deep neural models to capture temporal patterns and outputs second-order confidence to reflect the implicit stopping signals. Specifically, our model exploits the data not only from a time step but also from the probability sequence to capture stopping signals. By combining stopping signals from the classifier output and the second-order confidence, we design a more robust trigger to decide whether or not to request more observations from future time steps. Experimental results show that our approach can achieve superior results in early classification compared to state-of-the-art approaches.

Many but not all deep neural network audio models capture brain responses and exhibit correspondence between model stages and brain regions.

Models that predict brain responses to stimuli provide one measure of understanding of a sensory system and have many potential applications in science and engineering. Deep artificial neural networks have emerged as the leading such predictive models of the visual system but are less explored in audition. Prior work provided examples of audio-trained neural networks that produced good predictions of auditory cortical fMRI responses and exhibited correspondence between model stages and brain regions, but left it unclear whether these results generalize to other neural network models and, thus, how to further improve models in this domain. We evaluated model-brain correspondence for publicly available audio neural network models along with in-house models trained on 4 different tasks. Most tested models outpredicted standard spectromporal filter-bank models of auditory cortex and exhibited systematic model-brain correspondence: Middle stages best predicted primary auditory cortex, while deep stages best predicted non-primary cortex. However, some state-of-the-art models produced substantially worse brain predictions. Models trained to recognize speech in background noise produced better brain predictions than models trained to recognize speech in quiet, potentially because hearing in noise imposes constraints on biological auditory representations. The training task influenced the prediction quality for specific cortical tuning properties, with best overall predictions resulting from models trained on multiple tasks. The results generally support the promise of deep neural networks as models of audition, though they also indicate that current models do not explain auditory cortical responses in their entirety.

A Deep Multi-Tasking Approach Leveraging on Cited-Citing Paper Relationship For Citation Intent Classification

Citations are crucial artifacts to provide additional information to the reader to comprehend the research under concern. There are different roles that citations play in scientific discourse. Correctly identifying the intent of the citations finds applications ranging from predicting scholarly impact, finding idea propagation, to text summarization. With the rapid growth in scientific literature, the need for automated methods to classify citations is now growing intense. However, we can only fully understand the intent of a citation if we look at the citation context in the citing paper and also the primary purpose of the cited article. In this work, we propose a neural multi-task learning framework that harnesses the structural information of the research papers and the cited paper’s information for the effective classification of citation intents. We analyze the impact of three auxiliary tasks on the performance of our approach for citation classification. Our experiments on three benchmark citation classification datasets show that incorporating cited paper information (title) shows that our deep neural model achieves a new state-of-the-art on the ACL-ARC dataset with an absolute increase of 5.3% in the F1 score over the previous best model. We also achieve comparable performance with respect to the best-performing systems in the SDP 2021 3C Shared task on Citation Context Classification. We make our codes available at https://github.com/Tirthankar-Ghosal/citationclassification-SCIM