Deep Neural Network Algorithm Research Articles

Deep Learning–Based Estimation of Implantable Collamer Lens Vault Using Optical Coherence Tomography

To develop and validate a deep learning neural network for automated measurement of implantable collamer lens (ICL) vault using anterior segment optical coherence tomography (AS-OCT). Cross-sectional retrospective study. Multicentric. A total of 2,647 AS-OCT anonymized eye scans were used from 139 eyes of 82 subjects who underwent ICL surgery in 3 different centers. Using transfer learning, a deep learning neural network algorithm was trained and validated for the regression task of estimating the ICL vault on the MS-39 OCT (CSO, Florence, Italy). A trained operator separately reviewed all OCT scans and measured the central vault using a built-in OCT caliper tool. The model was then separately tested on 191 AS-OCT scans. A Bland-Altman plot was constructed and the mean absolute percentage error (MAPE), the mean absolute error (MAE), the root mean squared error (RMSE), the Pearson correlation coefficient (r), and the determination coefficient (R2) were calculated to evaluate the strength and validity of the model. On the test set, the model achieved a MAPE of 3.42%, an MAE of 15.82 µm, an RMSE of 18.85 µm, a Pearson correlation coefficient r of +0.98 (P-Value < .00001), and a coefficient of determination R2 of +0.96. There was no significant difference between the vaults of the test set labeled by the technician versus those estimated by the model: 478 ± 95 µm versus 475 ± 97 µm, respectively, P-Value = 0.064). Using transfer learning, our deep learning neural network was able to accurately compute the ICL vault from AS-OCT scans, overcoming the limitations of an imbalanced dataset and limited training data. Such an algorithm can assist the postoperative assessment in ICL surgery. PRéCIS: This study develops and validates a deep learning model to accurately estimate implantable collamer lens vault on optical coherence tomography using transfer learning, convoluted neural networks, and regression.

Numerical Study on Kinetic Effects of Driving Frequency in Atmospheric Radio Frequency Discharges Using Deep Neural Network

The particle-in-cell Monte Carlo collision (PIC-MCC) model is an essential way to investigate the kinetic behaviors of low-temperature plasmas, but usually, it is hugely time consuming in simulating atmospheric radio frequency (RF) plasmas. In this study, a deep neural network (DNN) with multiple hidden layers is developed to predict the kinetic discharge characteristics of atmospheric RF plasmas. The results obtained from the PIC-MCC model are used as the training dataset for the DNN, and the well-trained DNN is able to efficiently yield various kinetic behaviors of atmospheric RF discharges with very high precision. The validation of the results predicted by the DNN algorithm is performed by comparing them with the simulation results directly from the PIC-MCC model. Compared with the time-consuming PIC-MCC simulations, the well-trained DNN takes only 0.01 s to yield the essential kinetic characteristics of atmospheric RF discharges, which saves seven orders of magnitude of computation time compared with the traditional PIC-MCC simulation. The predicted data show that the discharge current of atmospheric RF discharges increases monotonically with the driving frequency at a given applied voltage, and as the driving frequency increases, the electric field in the sheath region is strongly enhanced with the sheath region shrinks and the bulk plasma region expanding. Additionally, the electron energy distribution function (EEDF) can be accurately predicted by DNN; moreover, as the driving frequency increases, the low-energy electrons can be transformed into medium-energy electrons, leading to a transition of the EEDF from a three-temperature distribution to a Maxwellian distribution from the DNN prediction. The study indicates that the well-trained DNN is a promising tool for plasma simulation with very high efficiency and accuracy compared with the PIC-MCC model with huge computational costs, which could provide enough kinetic results to further understand the discharge behaviors in atmospheric RF discharges.

Stochastic perturbation of subgradient algorithm for nonconvex deep neural networks

Stochastic perturbation of subgradient algorithm for nonconvex deep neural networks

Diagnosis of Cardiovascular Disease using Deep Learning

Abstract: Cardiovascular disease is one of the most horrendous illnesses, particularly the silent heart attack that strikes a person so abruptly that there is no time for treatment. It's difficult to diagnose a disease of this nature. One of the scariest diseases that c an kill a person at any time without warning is heart disease, and most doctors are unable to predict silent heart attacks. The lac k of specialists and an increase in cases of wrong diagnoses have fueled the demand for the creation of an efficient cardiovascul ar disease prediction system. This resulted in the exploration and development of original machine learning and medical data mi ning methodologies. The principal goal of this research is to identify the most crucial qualities for silent heart attack identificatio n by using classification algorithms to extract significant patterns and features from medical data. Although it is not innovative t o build such a system, the current ones have flaws and are not designed to detect the likelihood of silent heart attacks. Another is sue with the present heart attack prediction method is the use of characteristics. Choosing the typical features for the heart attac k prediction algorithm frequently yields unreliable results. To increase prediction accuracy, the suggested method aims to extract suitable attributes from the datasets. We developed a framework in this exploration that can understand the principles of predicti ng the risk profile of patients with the clinical data parameters. This research suggests an effective neural network with convolut ional layers to classify clinical data that is noticeably class-imbalanced. In order to forecast the development of Coronary Heart Disease, data from the National Health and Nutritional Examination Survey (NHANES) is collected (CHD). This research aime d to design a robust deep-learning algorithm to predict heart disease. Heart disease prediction is performed using SMOTE and MLP Classifier algorithms and Deep Neural Network Algorithms. The effectiveness of the model that accurately predicts the pre sence or absence of heart disease was examined using DNN and ANN. In this research article, we'll look at a machine-learnin g model that can clearly assess cardiac issues and be utilized by analysts and medical professionals

Large-Scale Road Network Traffic Congestion Prediction Based on Recurrent High-Resolution Network

Traffic congestion is a significant problem that adversely affects the economy, environment, and public health in urban areas worldwide. One promising solution is to forecast road-level congestion levels in the short-term and long-term, enabling commuters to avoid congested areas and allowing traffic agencies to take appropriate action. In this study, we propose a hybrid deep neural network algorithm based on High-Resolution Network (HRNet) and ConvLSTM decoder for 10, 30, and 60-min traffic congestion prediction. Our model utilizes the HRNet’s multi-scale feature extraction capability to capture rich spatial features from a sequence of past traffic input images. The ConvLSTM module learns temporal information from each HRNet multi-scale output and aggregates all feature maps to generate accurate traffic forecasts. Our experiments demonstrate that the proposed model can efficiently and effectively learn both spatial and temporal relationships for traffic congestion and outperforms four other state-of-the-art architectures (PredNet, UNet, ConvLSTM, and Autoencoder) in terms of accuracy, precision, and recall. A case study was conducted on the dataset from Seoul, South Korea.

Novel Cuckoo Search-Based Metaheuristic Approach for Deep Learning Prediction of Depression

Depression is a common illness worldwide with doubtless severe implications. Due to the absence of early identification and treatment for depression, millions of individuals worldwide suffer from mental illnesses. It might be difficult to identify those who are experiencing mental health illnesses and to provide them with the early help that they need. Additionally, depression may be associated with thoughts of suicide. Currently, there are no clinically specific diagnostic biomarkers that can identify the severity and type of depression. In this research paper, the novel particle swarm-cuckoo search (PS-CS) optimization algorithm is proposed instead of the traditional backpropagation algorithm for training deep neural networks. The backpropagation algorithm is widely used for supervised learning in deep neural networks, but it has limitations in terms of convergence speed and the possibility of getting trapped in local optima. These problems were addressed by using a deep neural network architecture for depression detection tasks along with the PS-CS optimization technique. The PS-CS algorithm combines the strengths of both particle swarm optimization and cuckoo search algorithms, which allows for a more efficient and effective optimization of the network parameters. We also evaluated how well the suggested methods performed against the most widely used classification models, including (K-nearest neighbor) KNN, (support vector regression) SVR, and decision trees, as well as the most widely used deep learning models, including residual neural network (ResNet), visual geometry group (VGG), and simple neural network (LeNet). The findings show that the suggested method, PS-CS, in conjunction with the CNN model, outperformed all other models, achieving the maximum accuracy of 99.5%. Other models, such as the KNN, decision trees, and logistic regression, achieved lower accuracies ranging from 69% to 97%.

Wrist-Based Electrodermal Activity Monitoring for Stress Detection Using Federated Learning.

With the most recent developments in wearable technology, the possibility of continually monitoring stress using various physiological factors has attracted much attention. By reducing the detrimental effects of chronic stress, early diagnosis of stress can enhance healthcare. Machine Learning (ML) models are trained for healthcare systems to track health status using adequate user data. Insufficient data is accessible, however, due to privacy concerns, making it challenging to use Artificial Intelligence (AI) models in the medical industry. This research aims to preserve the privacy of patient data while classifying wearable-based electrodermal activities. We propose a Federated Learning (FL) based approach using a Deep Neural Network (DNN) model. For experimentation, we use the Wearable Stress and Affect Detection (WESAD) dataset, which includes five data states: transient, baseline, stress, amusement, and meditation. We transform this raw dataset into a suitable form for the proposed methodology using the Synthetic Minority Oversampling Technique (SMOTE) and min-max normalization pre-processing methods. In the FL-based technique, the DNN algorithm is trained on the dataset individually after receiving model updates from two clients. To decrease the over-fitting effect, every client analyses the results three times. Accuracies, Precision, Recall, F1-scores, and Area Under the Receiver Operating Curve (AUROC) values are evaluated for each client. The experimental result shows the effectiveness of the federated learning-based technique on a DNN, reaching 86.82% accuracy while also providing privacy to the patient's data. Using the FL-based DNN model over a WESAD dataset improves the detection accuracy compared to the previous studies while also providing the privacy of patient data.

Automatic recognition of teeth and periodontal bone loss measurement in digital radiographs using deep-learning artificial intelligence

Background/purposeArtificial Intelligence (AI) can optimize treatment approaches in dental healthcare due to its high level of accuracy and wide range of applications. This study seeks to propose a new deep learning (DL) ensemble model based on deep Convolutional Neural Network (CNN) algorithms to predict tooth position, detect shape, detect remaining interproximal bone level, and detect radiographic bone loss (RBL) using periapical and bitewing radiographs. Materials and methods270 patients from January 2015 to December 2020, and all images were deidentified without private information for this study. A total of 8000 periapical radiographs with 27,964 teeth were included for our model. AI algorithms utilizing the YOLOv5 model and VIA labeling platform, including VGG-16 and U-Net architecture, were created as a novel ensemble model. Results of AI analysis were compared with clinicians' assessments. ResultsDL-trained ensemble model accuracy was approximately 90% for periapical radiographs. Accuracy for tooth position detection was 88.8%, tooth shape detection 86.3%, periodontal bone level detection 92.61% and radiographic bone loss detection 97.0%. AI models were superior to mean accuracy values from 76% to 78% when detection was performed by dentists. ConclusionThe proposed DL-trained ensemble model provides a critical cornerstone for radiographic detection and a valuable adjunct to periodontal diagnosis. High accuracy and reliability indicate model's strong potential to enhance clinical professional performance and build more efficient dental health services.

Deep learning inversion method of tunnel resistivity synthetic data based on modelling data

AbstractDifferent water‐bearing geological structures in front of the tunnel face are the main cause of tunnel water inrush disasters, affecting tunnel constructionsafety. Due to the narrow tunnel space and the limited data that can be detected, the traditional linear inversion method for detecting them has multiple solutions. In this paper, we establish a database with complex models, including water‐bearing geological structures usually encountered during tunnel construction. Then we build a complex nonlinear relationship mapping between the tunnel face observation data and the resistivity model through the deep neural network algorithm (electrical resistivity inversion network tunnel, ERTInvNet‐T for short). ERTInvNet‐T first uses a fully connected network to extract the tunnel depth features from the observed data, from which a three‐dimensional geoelectric model of the tunnel is then generated by a three‐dimensional deconvolutional network. At the same time, there is a problem with the uneven spatial distribution of the sensitivity of the data to the model. Therefore, depth weighting information constraint based on the distance factor is added to the loss function, which improves the network algorithm's learning ability for different detection positions of the tunnel. The validity of the proposed method is verified by a large number of numerical simulation experiments.

Predicting Ventilation Rate in a Naturally Ventilated Dairy Barn in Wind-Forced Conditions Using Machine Learning Techniques

Precise ventilation rate estimation of a naturally ventilated livestock building can benefit the control of the indoor environment. Machine learning has become a useful technique in many research fields and might be applied to ventilation rate prediction. This paper developed a machine-learning model for ventilation rate prediction from batch computational fluid dynamics (CFD) simulation results. By comparing deep neural networks (DNN), support vector regression (SVR), and random forest (RF), the best machine learning algorithm was selected. By comparing the modeling scheme of direct single-output (ventilation rate) and indirect multiple-output (predict averaged air velocities normal to the openings, then calculate the ventilation rate), the performances of the machine learning models widely applied in ventilation rate prediction were evaluated. In addition, this paper further evaluated the impact of adding indoor air velocity measurement in ventilation rate prediction. The results showed that the modeling performance of the DNN algorithm (Mean Absolute Percentage Error (MAPE) = 20.1%) was better than those of the SVR (MAPE = 23.2%) and RF algorithm (MAPE = 21.0%). The scheme of multiple-output performed better (MAPE &lt; 8%) than the single-output scheme (MAPE = 20.1%), where MAPE was the mean absolute percentage error. Additionally, the comparison of modeling schemes with different inputs showed that the predictive accuracy could be improved by adding indoor velocities to the inputs. The MAPE decreased from 7.7% in the scheme without indoor velocity to 4.4% in the scheme with one indoor velocity, and 3.1% in the scheme with two indoor velocities. The location of the additional air velocity affected the accuracy of the predictive model, with the ones at the bottom layer performing better in the prediction than those at the top layer. This study enables a real-time and accurate prediction of the ventilation rate of a barn and provides a recommendation for optimal indoor sensor placement.

Multi-level feature re-weighted fusion for the semantic segmentation of crops and weeds

Intelligent farm robots empowered by proper vision algorithms are the new agricultural machinery that eases weed control with speed and accuracy. Based on the farmland substantial similarity between the crops, and weeds, or other background interference objects, an improved deep convolutional neural network (DCNN) algorithms is proposed for the pixel semantic segmentation of crop and weed. First, a lightweight backbone is proposed to balance the features map textual and shape signals, which are essential cues for better crop and weed prediction. Second, a multi-level feature re-weighted fusion (MFRWF) module is suggested to combine only the relevant information from every backbone layer output to improve the contextual maps of crops and weeds. Finally, a decoder is designed based on convolutional weighted fusion (CWF) to preserve the relevant crop and weed context information by reducing the possible feature context distortion. Experimental results show that our improved neural network obtained the mean intersection of union (MIOU) scores of 0.8646, 0.9164, and 0.8459 on the carrot/weed field image (CWFID), sugar beet (BoniRob), and Rice seedling datasets, respectively. Therefore, the results have not only outperformed the commonly used architectures but can precisely identify crops/weeds and substantially improve the robot inference speed with minimal memory overhead. The code is available at:https://github.com/jannehlamin/MFRWF.

An Endodontic Forecasting Model Based on the Analysis of Preoperative Dental Radiographs: A Pilot Study on an Endodontic Predictive Deep Neural Network

IntroductionThis study aimed to evaluate the use of deep convolutional neural network (DCNN) algorithms to detect clinical features and predict the three-year outcome of endodontic treatment on preoperative periapical radiographs. MethodsA database of single-root premolars that received endodontic treatment or retreatment by endodontists with presence of three-year outcome was prepared (n = 598). We constructed a 17-layered DCNN with a self-attention layer (Periapical Radiograph Explanatory System with Self-Attention Network [PRESSAN-17]), and the model was trained, validated, and tested to 1) detect 7 clinical features, that is, full coverage restoration, presence of proximal teeth, coronal defect, root rest, canal visibility, previous root filling, and periapical radiolucency and 2) predict the three-year endodontic prognosis by analyzing preoperative periapical radiographs as an input. During the prognostication test, a conventional DCNN without a self-attention layer (residual neural network [RESNET]-18) was tested for comparison. Accuracy and area under the receiver-operating-characteristic curve were mainly evaluated for performance comparison. Gradient-weighted class activation mapping was used to visualize weighted heatmaps. ResultsPRESSAN-17 detected full coverage restoration (area under the receiver-operating-characteristic curve = 0.975), presence of proximal teeth (0.866), coronal defect (0.672), root rest (0.989), previous root filling (0.879), and periapical radiolucency (0.690) significantly, compared to the no-information rate (P < .05). Comparing the mean accuracy of 5-fold validation of 2 models, PRESSAN-17 (67.0%) showed a significant difference to RESNET-18 (63.4%, P < .05). Also, the area under average receiver-operating-characteristic of PRESSAN-17 was 0.638, which was significantly different compared to the no-information rate. Gradient-weighted class activation mapping demonstrated that PRESSAN-17 correctly identified clinical features. ConclusionsDeep convolutional neural networks can detect several clinical features in periapical radiographs accurately. Based on our findings, well-developed artificial intelligence can support clinical decisions related to endodontic treatments in dentists.

A machine learning approach for correcting radial velocities using physical observables

Context. Precision radial velocity (RV) measurements continue to be a key tool for detecting and characterising extrasolar planets. While instrumental precision keeps improving, stellar activity remains a barrier to obtaining reliable measurements below 1–2 m s−1 accuracy. Aims. Using simulations and real data, we investigate the capabilities of a deep neural network approach to producing activity-free Doppler measurements of stars. Methods. As case studies we used observations of two known stars, ϵ Eridani and AU Microscopii, both of which have clear signals of activity-induced Doppler variability. Synthetic observations using the starsim code were generated for the observables (inputs) and the resulting Doppler signal (labels), and then they were used to train a deep neural network algorithm to predict Doppler corrections. We identified a relatively simple architecture, consisting of convolutional layers followed by fully connected layers, that is adequate for the task. The indices investigated are mean line-profile parameters (width, bisector, and contrast) and multi-band photometry. Results. We demonstrate that the RV-independent approach can drastically reduce spurious Doppler variability from known physical effects, such as spots, rotation, and convective blueshift. We identify the combinations of activity indices with the most predictive power. When applied to real observations, we observe a good match of the correction with the observed variability, but we also find that the noise reduction is not as good as in the simulations, probably due to a lack of detail in the simulated physics. Conclusions. We demonstrate that a model-driven machine learning approach is sufficient to clean Doppler signals from activity-induced variability for well-known physical effects. There are dozens of known activity-related observables whose inversion power remains unexplored, indicating that the use of additional indicators, more complete models, and more observations with optimised sampling strategies can lead to significant improvements in our detrending capabilities for new and existing datasets.

FPGA Hardware Acceleration Design for Deep Learning

A type of artificial neural network called a convolutional neural network (CNN) can learn characteristics from a huge amount of data and performs very well in the field of large-scale image processing. CNN simulates the behavior of a biological optic nerve. In recent years, with the development of deep neural network algorithms and hardware technology, the current "CPU+GPU" model servers cannot meet the neural network structure in various fields, so a large amount of deep CNN accelerators based on the FPGA platform have gradually emerged. FPGA is beginning to be used in the fields of image recognition and natural language processing because of its programmability, high performance, high stability, high security, and low power consumption. Though FPGA has proven to have better performance, there is still room for optimization at the design level. Yolov3, as a classical algorithm, still consumes a lot of time and computational resources in actual operations. To address this problem, this experiment partially optimizes the Yolov3 algorithm by introducing the CBAM attention mechanism in the Yolov3 model and pruning the embedded system with different proportions using the Network slimming method. Finally, it is verified on a TX2 embedded device developed by Nvidia using the COCO dataset. The experiment finds that the precision, mAP, and the number of parameters of the optimized Yolov3 algorithm under different optimization strategies. It is shown that the Yolov3 algorithm still has more optimization strategies that can reduce the time required for computation and the memory occupied more effectively without any degradation in accuracy.

Construction of an efficient evaluation model for athletic athletes' competitive ability based on deep neural network algorithm

This paper analyzes the data of this year's athletes' physical fitness test scores and manages the classification of different physical qualities of the farmer. In order to reduce the manual calculation and increase the prediction efficiency, as well as to unify the scoring criteria of previous years, this paper proposes a comprehensive performance prediction model based on deep neural network algorithm. First, principal component analysis is used to transform multiple attributes with strong correlation into independent attributes that are not related to each other, and to reduce the time and space for model training by eliminating redundancy. Second, a back propagation (BP) neural network algorithm is used to build a physical fitness test prediction model, and the model is applied to the test dataset for model performance evaluation. Finally, the physical fitness test model was applied to other years for comprehensive performance prediction, and the differences between the model prediction results and the actual teachers' manual calculation results were observed. The results showed very good prediction results for 2021, in which 92.95% of the data had an absolute value of error less than 2 and only 0.06% had an absolute value of error greater than 4, which indicated that the prediction performance of the model was extremely significant. At the same time, a new athletic athletic scoring standard was also developed based on the neural network BP model to provide a more scientific theoretical basis and guidance for the evaluation of athletic ability of athletes.

4Gbaud PS-16QAM D-Band Fiber-Wireless Transmission over 4.6 km by Using Balance Complex-Valued NN Equalizer with Random Oversampling.

D-band (110-170 GHz) is a promising direction for the future of 6th generation mobile networks (6G) for high-speed mobile communication since it has a large available bandwidth, and it can provide a peak rate of hundreds of Gbit/s. Compared with the traditional electrical approach, photonics millimeter wave (mm-wave) generation in D-band is more practical and effectively overcomes the bottleneck of electrical devices. However, long-distance D-band wireless transmission is still limited by some key factors such as large absorption loss and nonlinear noises. Deep neural network algorithms are regarded as an important technique to model the nonlinear wireless behavior, among which the study on complex-value equalization is critical, especially in coherent detection systems. Moreover, probabilistic shaping is useful to improve the transmission capacity but also causes an imbalanced machine learning issue. In this paper, we propose a novel complex-valued neural network equalizer coupled with balanced random oversampling (ROS). Thanks to the adaptive deep learning method for probabilistic shaping-quadrature amplitude modulation (PS-QAM), we successfully realize a 135 GHz 4Gbaud PS-16QAM with a shaping entropy of 3.56 bit/symbol wireless transmission over 4.6 km. The bit error ratio (BER) of 4Gbaud PS-16QAM can be decreased to a soft-decision forward error correction (SD-FEC) with a 25% overhead of 2 × 10-2. Therefore, we can achieve a net rate of an 11.4 Gbit/s D-band radio-over-fiber (ROF) delivery over 4.6 km air free wireless distance.

Spatial distribution of fishing intensity of canvas stow net fishing vessels in the East China Sea and the Yellow Sea

Present study used the position data of BeiDou Vessel Monitoring System (VMS) in 2018, with respect to motorised fishingvessels in the East China Sea and the Yellow Sea to construct a fishing vessel operating status classification model based onthreshold, deep neural network and DBSCAN density clustering algorithm. The geographic grid was divided into cells of0.1°×0.1° and the average fishing time per square km (h km-2) in each grid was calculated to obtain the spatial distributionof fishing intensity in the study region in 2018. The results showed that the threshold method could classify fishing vesselsailing, anchoring and other states with an accuracy of more than 95%. The deep neural network and DBSCAN algorithmcould classify the two states of netting and closing with an accuracy of 94.7%. By classifying the status of fishing vessels,quantitative monitoring can be carried out to better serve the management of marine fishery resources and marine ecologicalprotectionKeywords: China, DBSCAN, Deep neural network, Fishing intensity, Spatial distribution, VMS, Voyage extraction

Short Text Sentiment Classification Using Bayesian and Deep Neural Networks

The previous multi-layer learning network is easy to fall into local extreme points in supervised learning. If the training samples sufficiently cover future samples, the learned multi-layer weights can be well used to predict new test samples. This paper mainly studies the research and analysis of machine short text sentiment classification based on Bayesian network and deep neural network algorithm. It first introduces Bayesian network and deep neural network algorithms, and analyzes the comments of various social software such as Twitter, Weibo, and other popular emotional communication platforms. Using modeling technology popular reviews are designed to conduct classification research on unigrams, bigrams, parts of speech, dependency labels, and triplet dependencies. The results show that the range of its classification accuracy is the smallest as 0.8116 and the largest as 0.87. These values are obtained when the input nodes of the triple dependency feature are 12,000, and the reconstruction error range of the Boltzmann machine is limited between 7.3175 and 26.5429, and the average classification accuracy is 0.8301. The advantages of triplet dependency features for text representation in text sentiment classification tasks are illustrated. It shows that Bayesian and deep neural network show good advantages in short text emotion classification.

Hierarchical optimization method of space patrol task assignment based on deep neural network and consensus-based bundle algorithm

This research looks towards improving the accuracy and efficiency of orbit transfer speed increment estimation and task assignment for space patrol missions involving multiple platforms. First, we set up a quick prediction model for the increment in orbit transfer velocity. Then, a two-layered hybrid optimization framework is developed, with a deep neural network sub-model as the inner layer to find the optimal speed increase and the consensus-based bundle algorithm as the outer layer to solve the multi-task patrol sequence problem. Finally, the simulation and analysis of the patrol task assignment problem are conducted under two types of scenarios. The results show that the hierarchical optimization method proposed in this paper is three orders of magnitude faster than the traditional optimization method. In addition, the suggested hierarchical optimization method has significant promise in practical applications, where it can produce solutions more quickly to adapt to more complex patrol scenarios.

The Optimization Strategy of Deep Learning Algorithm Based on Image Recognition and Object Detection

With the advent of intelligent era, the application of artificial intelligence has penetrated all walks of life in our daily lives and has gradually entered the scope of being understood by the public, such as AlphaGo, which is a human-machine game in Go. Meanwhile, with the increasing complexity of deep neural network algorithms, how to compress and optimize neural network models becomes the key to reducing model storage space and improving model deployment efficiency. This article will complete the INT8 quantization of the model for AlexNet and use the concept of batch normalization (BN) layer to prune the Yolov3 model. Finally, it is proved by experiments that the model quantization makes the image recognition time of the AlexNet about 1/3-1/4 of the original format; and pruning of Yolov3 will significantly reduce the storage space occupied by the model, moreover, make the speed doubled of embedded GPU real-time object detection. However, none of the three experiments reduces the model recognition accuracy. This paper proposes some essential concepts in the main body and applies them to model algorithms and experiments to provide theoretical support and proof for the field of deep learning model compression and optimization. Furthermore, it is possible to explore cutting-edge research directions like Sparse Convolution Net based on the research methods and experimental conclusions of the paper.