Convolutional Neural Network Framework Research Articles

A Stock Trading Expert System Established by the CNN-GA-Based Collaborative System

This article uses a new convolutional neural network framework, which has good performance for time series feature extraction and stock price prediction. This method is called the stock sequence array convolutional neural network, or SSACNN for short. SSACNN collects data on leading indicators including historical prices and their futures and options, and uses arrays as the input map of the CNN framework. In the financial market, every number has its logic behind it. Leading indicators such as futures and options can reflect changes in many markets, such as the industry's prosperity. Adding the data set of leading indicators can predict the trend of stock prices well. This study takes the stock markets of the United States and Taiwan as the research objects and uses historical data, futures, and options as data sets to predict the stock prices of these two markets, and then uses genetic algorithms to find trading signals, so as to get a stock trading system. The experimental results show that the stock trading system proposed in this research can help investors obtain certain returns.

An Optimized Convolutional Neural Network for Multi-Spectral Change Detection Technique

With the proliferation of multi-sensor remote sensing images, extraction of information has progressively concerned attention in recent years. It leads to great significant improvement in remote sensing areas for change detection techniques. Due to the advent of multi-spectral images in the change detection model, a change detection model named optimized convolutional neural network (CNN) framework is designed for monitoring land cover changes. This can assist to note the changes happening in the different periods. The proposed model is designed by integrating the normalization function and detection model by applying the deep learning (DL) algorithm. The actual experiment is carried out by taking the parameters accuracy and Kappa coefficient to show the effectiveness of the proposed model. The proposed experiment out performed well when compared to the existing mainstream techniques in multi-sensor images change detection. When compared with the existing DL techniques, the proposed design attained accuracy of 98.92% by possessing less loss function.

Transfer learning using small-sized dataset for concentrate ash content prediction of coal flotation

ABSTRACT Real-time and accurate acquisition of ash content is crucial for coal flotation monitoring but existing modeling approaches have been hindered from a large dataset labeled by the target task. The method proposed in this paper relies on a transfer learning approach that is able to exploit the knowledge of a source domain (ImageNet and S1) to improve the result of a task performed on the target domain (S2). The ResNet 101 is selected as the main Convolutional neural network (CNN) framework of transfer learning among six state-of-the-art CNN modules in this study. Freeze and fine-tune are used to transfer the weights of the CNN layers to the industry ash content classification task. The tuning results of neurons and fully connected layers show that the classification results based on the weights knowledge of S1 obtained through the proposed transfer method outperform the training direct method by achieving an accuracy of 86.26% and F1-score of 85.19% over the small-sized industry dataset. The 3-D discriminant feature score and ROC curves demonstrate satisfactory prediction performance in coal flotation industrial ash content prediction. Generally, this work provides new options for flotation monitoring in tasks of training ash content labels shortage.

Privacy‐preserving CNN feature extraction and retrieval over medical images

Online medicine diagnosis based on pathological images has been regarded as a pervasive method due to the advances in electronic healthcare and Internet of Things (IoT), however, it also causes storage and computing stress on the local IoT devices. To solve this problem, a nature way is to outsource images to cloud servers. Unfortunately, a range of security and privacy issues arise while delegating both storage and computing to the untrusted external servers. In this paper, we present a privacy-preserving feature extraction and retrieval scheme over medical images, which allows images storage and processing on two separate cloud servers. We share the images using secret sharing technology and design a set of secure two-party computation protocols between the two cloud servers. Then a privacy-preserving convolutional neural networks (CNN) framework is constructed to achieve feature extraction, classification and retrieval of images in the encrypted domain. We analyse and evaluate our scheme in terms of both security and efficiency. The results indicate that the proposed secure protocols in our scheme can significantly reduce the computation overhead while protecting the privacy of images as well as the data generated during execution on cloud servers and final results. The performance of our scheme in image feature extraction, classification and retrieval is at a similar level comparable to the scheme based on original CNN in plaintext.

CRISPR-OTE: Prediction of CRISPR On-Target Efficiency Based on Multi-Dimensional Feature Fusion

ObjectiveClustered Regularly Interspaced Short Palindromic Repeats (CRISPR) is a powerful genome editing technology. Guide RNA (gRNA) plays an essential guiding role in the CRISPR system by complementary base pairing with target DNA. Since the CRISPR targeting mechanism problem has not yet been fully resolved, it remains a challenge to predict gRNA on-target efficiency. Current gRNA design tools often lack efficient information extraction and cannot learn the target efficiency patterns thoroughly. Material and methodsIn this study, CRISPR-OTE is proposed to consider both multi-dimensional sequence information and important complementary prior knowledge based on a simple but effective framework. CRISPR-OTE consists of the local-contextual information branch and the prior knowledge branch. The local-contextual information branch extracts multi-dimensional sequence features from the DNA primary sequence by a parallel framework of Convolutional Neural Networks (CNN) and bidirectional Long Short-Term Memory networks (biLSTM). The prior knowledge branch selects the optimal subset of physicochemical features to provide the neural network with complementary knowledge, such as complex secondary structures. A simple feature fusion strategy is also adopted to fully utilize multi-modal data from the two branches. ResultsThe experimental results show that the optimal subset of physicochemical features (RNA secondary structure and melting temperature of 34nt target) can effectively improve the prediction performance. Additionally, combining multi-dimensional sequence features and multi-modal features can extract information more comprehensively. Through transfer learning, CRISPR-OTE trained on the CRISPR-Cpf1 system can also be successfully applied to the CRISPR-Cas9 system. ConclusionThe performance of CRISPR-OTE is superior to other methods in different CRISPR systems and species. Therefore, CRISPR-OTE is a simple on-target efficiency prediction framework with better accuracy and generalization performance.

A privacy preservation framework for feedforward-designed convolutional neural networks

A feedforward-designed convolutional neural network (FF-CNN) is an interpretable neural network with low training complexity. Unlike a neural network trained using backpropagation (BP) algorithms and optimizers (e.g., stochastic gradient descent (SGD) and Adam), a FF-CNN obtains the model parameters in one feed-forward calculation based on two methods of data statistics: subspace approximation with adjusted bias and least squares regression. Currently, models based on FF-CNN training methods have achieved outstanding performance in the fields of image classification and point cloud data processing. In this study, we analyze and verify that there is a risk of user privacy leakage during the training process of FF-CNN and existing privacy-preserving methods for model gradients or loss functions do not apply to FF-CNN models. Therefore, we propose a securely forward-designed convolutional neural network algorithm (SFF-CNN) to protect the privacy and security of data providers for the FF-CNN model. Firstly, we propose the DPSaab algorithm to add the corresponding noise to the one-stage Saab transform in the FF-CNN design for improved protection performance. Secondly, because noise addition brings the risk of model over-fitting and further increases the possibility of privacy leakage, we propose the SJS algorithm to filter the input features of the fully connected model layer. Finally, we theoretically prove that the proposed algorithm satisfies differential privacy and experimentally demonstrate that the proposed algorithm has strong privacy protection. The proposed algorithm outperforms the compared deep learning privacy-preserving algorithms in terms of utility and robustness.

Coupled Attention Framework of Convolutional Neural Network Based on Computer Intelligence.

Using an attention mechanism based on the convolutional neural networks (CNNs) improves the performance of computer vision tasks by enhancing the representation of the features. The existing attention methods enhance the expression of the features by modeling the internal information of the features. However, due to the limited information flow of the previous features, these methods are difficult to calibrate features more completely. In this paper, we propose a Coupled Attention Framework (CAF) that is a simple attention framework for improving the performance of the existing attention methods. In the CAF, a coupling branch is added to an existing attention method to generate the input attention maps and enhance the input features of the convolution. The input attention is then spread to the output features through coupling between the input attention maps and convolution, the output features. The final result is the experimental results on various visual tasks. The results show that applying CAF to most of the existing attention methods can improve the performance with fewer parameters.

Design of a quantum convolutional neural network on quantum circuits

This paper introduces a quantum convolutional neural network model that is implementable on real quantum circuits. Three basic blocks, including the quantum encoding block, the model design block and the parameter tuning block, are designed to formulate the quantum convolutional neural network framework. Simulation results on the MNIST dataset are presented to demonstrate the representation, learning, and optimization capacity of the proposed model.

BUILDINGS CHANGE DETECTION USING HIGH-RESOLUTION REMOTE SENSING IMAGES WITH SELF-ATTENTION KNOWLEDGE DISTILLATION AND MULTISCALE CHANGE-AWARE MODULE

Abstract. Building change detection from remote sensing images is vital for many applications such as urban planning and dynamic monitoring, smart city construction, and geographical information census. In recent years, with the improvement of artificial intelligence and computer vision techniques, deep learning algorithms, especially convolutional neural networks (CNN), provide automatic detection and extraction methods. Unlike traditional approaches relying on shallow manual features, CNN can generate the deep semantic features by fusing spatial and spectral information, which is conducive to identifying building change regions. However, most deep CNN models directly fuse different level features and recover spatial details, which probably introduce redundant background information and noise from shallow layers. Considering the building’s multiscale, convolution operation with fixed receptive fields cannot obtain a strong global feature response to changed regions. To address the above problems, we develop a CNN framework for automatic building region change detection using dual-temporal high-resolution remote sensing images. To refine the shallow features, self-attention knowledge distillation strategies are introduced to FCN. Furthermore, we propose the multiscale feature change-aware module to integrate the globally changed information in different decoders. Finally, the model aggregates the multi-scale regional change information and outputs the prediction map. The results of comparative visual analysis and quantitative evaluation in two public datasets demonstrate that the proposed network model can improve the accuracy and efficiency of the automatic building change detection (85.69 IoU, 97.56 OA on the WHU dataset, and 83.72 IoU, 97.64 OA on the LEVIR-CD dataset).

BFF R-CNN: Balanced Feature Fusion for Object Detection

In the neck part of a two-stage object detection network, feature fusion is generally carried out in either a top-down or bottom-up manner. However, two types of imbalance may exist: feature imbalance in the neck of the model and gradient imbalance in the region of interest extraction layer due to the scale changes of objects. The deeper the network is, the more abstract the learned features are, that is to say, more semantic information can be extracted. However, the extracted image background, spatial location, and other resolution information are less. In contrast, the shallow part can learn little semantic information, but a lot of spatial location information. We propose the Both Ends to Centre to Multiple Layers (BEtM) feature fusion method to solve the feature imbalance problem in the neck and a Multi-level Region of Interest Feature Extraction (MRoIE) layer to solve the gradient imbalance problem. In combination with the Region-based Convolutional Neural Network (R-CNN) framework, our Balanced Feature Fusion (BFF) method offers significantly improved network performance compared with the Faster R-CNN architecture. On the MS COCO 2017 dataset, it achieves an average precision (AP) that is 1.9 points and 3.2 points higher than those of the Feature Pyramid Network (FPN) Faster R-CNN framework and the Generic Region of Interest Extractor (GRoIE) framework, respectively.

ThreatNet: advanced threat detection, region-based convolutional neural network framework

It is critical for many countries to ensure public safety in detecting and identifying threats in a night, <span lang="EN-US">commercial places, border areas and public places. Majority of past research in this area has focused on the use of image-level categorization and object-level detection techniques. As an X-ray and thermal security image analysis strategy, object separation can considerably improve automatic threat detection when used in conjunction with other techniques. In order to detect possible threats, the effects of introducing segmentation deep learning models into the threat detection pipeline of a large imbalanced X-ray and thermal dataset were investigated. With the purpose of boosting the number of true positives discovered, a faster regional convolutional neural network (R-CNN) model was trained on a balanced dataset to identify probable hazard zones in X-ray and thermal security pictures. In order to get the final results, we combined the two models i.e faster R-CNN with Mask RCNN into a single detection pipeline using the transfer learning technique, which outperforms baseline and end-to-end instance segmentation methods using less number of the practical dataset, with mAPs ranging from 94.88 percent to 91.40 percent helps in detecting the person with guns, knives, pliers to avoid cross border threats.</span>

Automatic Detection of White Blood Cancer and Lung Cancer Using Machine Learning

Abstract: When abnormal white blood cells multiply cancerously, leukaemia results, harming the bone marrow and blood. The body's immune system is disturbed by the body's excessive generation of immature white blood cells, or lymphoblast, and its capacity to create red blood cells and platelets is decreased in the bone marrow as a result. Early diagnosis of these cells is crucial for preventing health issues in patients. In the past, the procedure was manually completed by an expert in a sizable amount of time. On a short dataset, the proposed model uses an improved Dense Convolutional neural network framework to predict the kind of cancer in a generic manner. which is able to locate both lung cancer and leukemia. Tumors can spread to other organs and grow quickly in lung cancer. An aberrant cell proliferation that can harm other normal tissue cells marks the beginning of cancer. Using tiny images to identify the white blood cancer, the imaging technology known as computerized tomography (CT) is frequently used to diagnose lung cancer.

Modelling monthly pan evaporation utilising Random Forest and deep learning algorithms

Evaporation is the primary aspect causing water loss in the hydrological cycle; therefore, water loss must be precisely measured. Evaporation is an intricate nonlinear process occurring as a result of several climatic aspects. The purpose of this research is to assess the feasibility of using Random Forest (RF) and two deep learning techniques, namely convolutional neural network (CNN), and deep neural network (DNN) to accurately estimate monthly pan evaporation rates. Month-based weather data gathered from four Malaysian weather stations during the 2000–2019 timeframe was used to train and evaluate the models. Several input attributes (predictor variables) were investigated to select the most suitable variables for machine learning models. Every approach was tested with several models, each with a different set of model aspects and input parameter combinations. The formulated ML approaches were benchmarked against two commonly used empirical methods: Stephens & Stewart and Thornthwaite. Model outcomes were assessed using standard statistical measures to determine their effectiveness in predicting evaporation. The results indicated that the three ML models developed in the study performed better than empirical models and could significantly improve the precision of monthly Ep estimates even with the identical input sets. The performance assessment metrics also show that the formulated CNN approach was acceptable for modelling monthly water loss due to evaporation with a higher degree of accuracy than other ML frameworks explored in this study. In addition, the CNN framework outperformed other AI techniques evaluated for the same areas using identical data inputs. The investigation’s findings in relation to the various performance criteria show that the proposed CNN model is capable of capturing the highly non-linearity of evaporation and could be regarded as an effective tool to predict evaporation.

F2F-NN: A Field-to-Field Wind Speed Retrieval Method of Microwave Radiometer Data Based on Deep Learning

In this paper, we present a method for retrieving sea surface wind speed (SSWS) from Fengyun-3D (FY-3D) microwave radiation imager (MWRI) data. In contrast to the conventional point-to-point (P2P) retrieval methods, we propose a field-to-field (F2F) SSWS retrieval method based on the basic framework of a Convolutional Neural Network (CNN). Considering the spatial continuity and consistency characteristics of wind fields within a certain range, we construct the model based on the basic framework of CNN, which is suitable for retrieving various wind speed intervals, and then synchronously obtaining the smooth and continuous wind field. The retrieval results show that: (1) Comparing the retrieval results with the label data, the root-mean-square error (RMSE) of wind speed is about 0.26 m/s, the F2F-NN model is highly efficient in training and has a strong fitting ability to label data. Comparing the retrieval results with the buoys (NDBC and TAO) data, the RMSE of F2F-NN wind speed is less than 0.91 m/s, the retrieval accuracy is better than the wind field products involved in the comparison. (2) In the hurricane (Sam) area, the F2F-NN model greatly improves the accuracy of wind speed in the FY-3D wind field. Comparing five wind field products with the Stepped-Frequency Microwave Radiometer (SFMR) data, the overall accuracy of the F2F-NN wind data is the highest. Comparing the five wind field products with the International Best Track Archive for Climate Stewardship (IBTrACS) data, the F2F-NN wind field is superior to the other products in terms of maximum wind speed and maximum wind speed radius. The structure of the wind field retrieved by F2F-NN is complete and accurate, and the wind speed changes smoothly and continuously.

3D Convolutional Neural Network Framework with Deep Learning for Nuclear Medicine.

Though artificial intelligence (AI) has been used in nuclear medicine for more than 50 years, more progress has been made in deep learning (DL) and machine learning (ML), which have driven the development of new AI abilities in the field. ANNs are used in both deep learning and machine learning in nuclear medicine. Alternatively, if 3D convolutional neural network (CNN) is used, the inputs may be the actual images that are being analyzed, rather than a set of inputs. In nuclear medicine, artificial intelligence reimagines and reengineers the field's therapeutic and scientific capabilities. Understanding the concepts of 3D CNN and U-Net in the context of nuclear medicine provides for a deeper engagement with clinical and research applications, as well as the ability to troubleshoot problems when they emerge. Business analytics, risk assessment, quality assurance, and basic classifications are all examples of simple ML applications. General nuclear medicine, SPECT, PET, MRI, and CT may benefit from more advanced DL applications for classification, detection, localization, segmentation, quantification, and radiomic feature extraction utilizing 3D CNNs. An ANN may be used to analyze a small dataset at the same time as traditional statistical methods, as well as bigger datasets. Nuclear medicine's clinical and research practices have been largely unaffected by the introduction of artificial intelligence (AI). Clinical and research landscapes have been fundamentally altered by the advent of 3D CNN and U-Net applications. Nuclear medicine professionals must now have at least an elementary understanding of AI principles such as neural networks (ANNs) and convolutional neural networks (CNNs).

Hybrid Beamforming for MISO System via Convolutional Neural Network

Hybrid beamforming (HBF) is a promising approach to obtain a better balance between hardware complexity and system performance in massive MIMO communication systems. However, the HBF optimization problem is a challenging task due to its nonconvex property in terms of design complexity and spectral efficiency (SE) performance. In this work, a low-complexity convolutional neural network (CNN)-based HBF algorithm is proposed to solve the SE maximization problem under the constant modulus constraint and transmit power constraint in a multiple-input single-output (MISO) system. The proposed CNN framework uses multiple convolutional blocks to extract more channel features. Considering that the solutions for the HBF are hard to obtain, we derive an unsupervised learning mechanism to avoid any labeled data when training the constructed CNN. We discuss the performance of the proposed algorithm in terms of both the generalization ability for multiple CSIs and the specific solving ability for an individual CSI, respectively. Simulations show its advantages in both SE and complexity over other related algorithms.

A 3D Convolutional Neural Network with Gradient Guidance for Image Super-Resolution of Late Gadolinium Enhanced Cardiac MRI.

In this paper, we describe a 3D convolutional neural network (CNN) framework to compute and generate super-resolution late gadolinium enhanced (LGE) cardiac magnetic resonance imaging (MRI) images. The proposed CNN framework consists of two branches: a super-resolution branch with a 3D dense deep back-projection network (DBPN) as the backbone to learn the mapping of low-resolution LGE cardiac volumes to high-resolution LGE cardiac volumes, and a gradient branch that learns the mapping of the gradient map of low resolution LGE cardiac volumes to the gradient map of their high-resolution counterparts. The gradient branch of the CNN provides additional cardiac structure information to the super-resolution branch to generate structurally more accurate super-resolution LGE MRI images. We conducted our experiments on the 2018 atrial segmentation challenge dataset. The proposed CNN framework achieved a mean peak signal-to-noise ratio (PSNR) of 30.91 and 25.66 and a mean structural similarity index measure (SSIM) of 0.91 and 0.75 on training the model on low-resolution images downsamp led by a scale factor of 2 and 4, respectively.

Learning-to-augment incorporated noise-robust deep CNN for detection of COVID-19 in noisy X-ray images

Deep convolutional neural networks (CNNs) are used for the detection of COVID-19 in X-ray images. The detection performance of deep CNNs may be reduced by noisy X-ray images. To improve the robustness of a deep CNN against impulse noise, we propose a novel CNN approach using adaptive convolution, with the aim to ameliorate COVID-19 detection in noisy X-ray images without requiring any preprocessing for noise removal. This approach includes an impulse noise-map layer, an adaptive resizing layer, and an adaptive convolution layer to the conventional CNN framework. We also used a learning-to-augment strategy using noisy X-ray images to improve the generalization of a deep CNN. We have collected a dataset of 2093 chest X-ray images including COVID-19 (452 images), non-COVID pneumonia (621 images), and healthy ones (1020 images). The architecture of pre-trained networks such as SqueezeNet, GoogleNet, MobileNetv2, ResNet18, ResNet50, ShuffleNet, and EfficientNetb0 has been modified to increase their robustness to impulse noise. Validation on the noisy X-ray images using the proposed noise-robust layers and learning-to-augment strategy-incorporated ResNet50 showed 2% better classification accuracy compared with state-of-the-art method.

Development and assessment of deep learning system for the location and classification of rib fractures via computed tomography

PurposeThe purpose of this study was to evaluate the performance of a deep learning system for the automatic diagnosis and classification of rib fractures. MethodsThis retrospective study analyzed computed tomography (CT) data of patients diagnosed with a rib fracture between 1 January 2019 and 23 July 2020 in two hospitals, including 591 patients from Suzhou TCM hospital and 75 patients from Jintan TCM hospital. A deep learning system (Dr.Wise@ChestFracture v1.0) based on a convolutional neural network framework was used as a diagnostic tool, and a human–model comparison experiment was designed to compare the diagnostic efficiencies of the deep learning system and radiologists. Furthermore, a secondary classification model was established to distinguish the different types of fracture. First, a classification model to differentiate between fresh and old fractures was developed. Second, a submodel to determine any misalignment in fresh fractures was established. ResultsFor all fracture types, the detection efficiency (recall) of the system was statistically significantly better than that of radiologists with different levels of experience (all p < 0.0167 except for senior radiologists). The F1-score of the system for diagnosing rib fractures was similar to that of the radiologists. The system was much faster than the radiologists in assessing rib fractures (all p < 0.0167). The two classification models can distinguish between fresh and old fractures (accuracy = 87.63%) and determine whether there is any misalignment in fresh fractures (accuracy = 95.22%) or not. ConclusionThe use of a deep learning system can accurately, automatically, and rapidly diagnose and classify rib fractures, helping doctors improve the diagnostic efficiency and reducing their workload. The classification models can distinguish different types of rib fracture well.

Research on three-dimensional counting algorithm of high-precision suspended cells based on deep learning

The research on the three-dimensional counting algorithm of suspension cells is very important for the online counting of cells. Suspension cells are usually observed and counted by taking out cell samples under the microscope. This method cannot realize the online counting of cells. At the same time, there will be weak movement of suspended cells in three-dimensional space, making space counting more difficult. Aiming at the above difficulties, a three-dimensional counting algorithm of suspended cells based on deep learning is proposed. The Faster regions with convolutional neural network (R-CNN) framework is used to identify the target cells, the feature pyramid structure is introduced to reduce the feature loss in the fusion process, and the soft non-maximum suppression algorithm is used to remove the bounding box to improve the detection effect. To reduce the manual labeling workload, an automatic labeling method is proposed. The improved Faster RCNN framework achieves the F1-score of more than 98% after identifying cells, and our method is more accurate than other advanced methods. In terms of three-dimensional counting algorithms, the cells are spatially tracked. Cell matching and correction are carried out by extracting multiple feature information of cells, and adhesion cells are extracted separately for priority processing. The algorithm carried out a tracking experiment on the image sequences of five groups of suspended cells, and the final tracking accuracy of the experiment reached more than 95%. The tracking accuracy is better than other advanced algorithms. Experimental results show that the algorithm achieves high-precision three-dimensional tracking and counting of suspended cells.