Deep Classification Research Articles

Unified adaptive deep classification for industrial real‐time situation awareness in edge environment

SummaryThe rapidly growing information technologies such as cloud computing, edge computing, artificial intelligence offer cutting‐edge technical means for virtual reality of industrial manufacturing process. An efficient and accurate method for situation awareness (SA) to monitor the process of marginal layer plays an essential role to enhance the safety of industrial equipment. In this study, a manufacturing life cycle is described and characterized as operation and remote controlling of industrial machineries. The edge servers are first employed to track physical devices. Then a novel unified adaptive deep classification framework for SA with semi‐supervised classification method is developed to find unknown patterns in a huge amount of industrial data. Furthermore, a robust model based on the selection of adaptive local neighbor is proposed to find out an optimal weight and high similarity of neighbors. The combination of the proposed framework and model can realize real‐time SA of modern manufacturing the execution system. Plentiful experiments have been conducted to demonstrate the high accuracy and reliability of the proposed framework.

Morphological screening of mesenchymal mammary tumor organoids to identify drugs that reverse epithelial-mesenchymal transition

The epithelial-mesenchymal transition (EMT) has been implicated in conferring stem cell properties and therapeutic resistance to cancer cells. Therefore, identification of drugs that can reprogram EMT may provide new therapeutic strategies. Here, we report that cells derived from claudin-low mammary tumors, a mesenchymal subtype of triple-negative breast cancer, exhibit a distinctive organoid structure with extended “spikes” in 3D matrices. Upon a miR-200 induced mesenchymal-epithelial transition (MET), the organoids switch to a smoother round morphology. Based on these observations, we developed a morphological screening method with accompanying analytical pipelines that leverage deep neural networks and nearest neighborhood classification to screen for EMT-reversing drugs. Through screening of a targeted epigenetic drug library, we identified multiple class I HDAC inhibitors and Bromodomain inhibitors that reverse EMT. These data support the use of morphological screening of mesenchymal mammary tumor organoids as a platform to identify drugs that reverse EMT.

Cluster-based Deep One-Class Classification Model for Anomaly Detection

Cluster-based Deep One-Class Classification Model for Anomaly Detection

RETRACTED: Breast cancer diagnosis using multiple activation deep neural network

Deep Neural Network (DNN) stands for multilayered Neural Network (NN) that is capable of progressively learn the more abstract and composite representations of the raw features of the input data received, with no need for any feature engineering. They are advanced NNs having repetitious hidden layers between the initial input and the final layer. The working principle of such a standard deep classifier is based on a hierarchy formed by the composition of linear functions and a defined nonlinear Activation Function (AF). It remains uncertain (not clear) how the DNN classifier can function so well. But it is clear from many studies that within DNN, the AF choice has a notable impact on the kinetics of training and the success of tasks. In the past few years, different AFs have been formulated. The choice of AF is still an area of active study. Hence, in this study, a novel deep Feed forward NN model with four AFs has been proposed for breast cancer classification: hidden layer 1: Swish, hidden layer, 2:-LeakyReLU, hidden layer 3: ReLU, and final output layer: naturally Sigmoidal. The purpose of the study is twofold. Firstly, this study is a step toward a more profound understanding of DNN with layer-wise different AFs. Secondly, research is also aimed to explore better DNN-based systems to build predictive models for breast cancer data with improved accuracy. Therefore, the benchmark UCI dataset WDBC was used for the validation of the framework and evaluated using a ten-fold CV method and various performance indicators. Multiple simulations and outcomes of the experimentations have shown that the proposed solution performs in a better way than the Sigmoid, ReLU, and LeakyReLU and Swish activation DNN in terms of different parameters. This analysis contributes to producing an expert and precise clinical dataset classification method for breast cancer. Furthermore, the model also achieved improved performance compared to many established state-of-the-art algorithms/models.

The effect of hyperparameter search on artificial neural network in human activity recognition

Abstract In the last decade, many researchers applied shallow and deep networks for human activity recognition (HAR). Currently, the trending research line in HAR is applying deep learning to extract features and classify activities from raw data. However, we observed that, authors of previous studies have not performed an efficient hyperparameter search on their artificial neural network (shallow or deep)-based classifier. Therefore, in this article, we demonstrate the effect of the random and Bayesian parameter search on a shallow neural network using five HAR databases. The result of this work shows that a shallow neural network with correct parameter optimization can achieve similar or even better recognition accuracy than the previous best deep classifier(s) on all databases. In addition, we draw conclusions about the advantages and disadvantages of the two hyperparameter search techniques according to the results.

COVID-19 deep classification network based on convolution and deconvolution local enhancement

Computer Tomography (CT) detection can effectively overcome the problems of traditional detection of Corona Virus Disease 2019 (COVID-19), such as lagging detection results and wrong diagnosis results, which lead to the increase of disease infection rate and prevalence rate. The novel coronavirus pneumonia is a significant difference between the positive and negative patients with asymptomatic infections. To effectively improve the accuracy of doctors' manual judgment of positive and negative COVID-19, this paper proposes a deep classification network model of the novel coronavirus pneumonia based on convolution and deconvolution local enhancement. Through convolution and deconvolution operation, the contrast between the local lesion region and the abdominal cavity of COVID-19 is enhanced. Besides, the middle-level features that can effectively distinguish the image types are obtained. By transforming the novel coronavirus detection problem into the region of interest (ROI) feature classification problem, it can effectively determine whether the feature vector in each feature channel contains the image features of COVID-19. This paper uses an open-source COVID-CT dataset provided by Petuum researchers from the University of California, San Diego, which is collected from 143 novel coronavirus pneumonia patients and the corresponding features are preserved. The complete dataset (including original image and enhanced image) contains 1460 images. Among them, 1022 (70%) and 438 (30%) are used to train and test the performance of the proposed model, respectively. The proposed model verifies the classification precision in different convolution layers and learning rates. Besides, it is compared with most state-of-the-art models. It is found that the proposed algorithm has good classification performance. The corresponding sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and precision are 0.98, 0.96, 0.98, and 0.97, respectively.

Deep Transfer Learning for Improved Detection of Keratoconus using Corneal Topographic Maps

Clinical keratoconus (KCN) detection is a challenging and time-consuming task. In the diagnosis process, ophthalmologists must revise demographic and clinical ophthalmic examinations. The latter include slit-lamb, corneal topographic maps, and Pentacam indices (PI). We propose an Ensemble of Deep Transfer Learning (EDTL) based on corneal topographic maps. We consider four pretrained networks, SqueezeNet (SqN), AlexNet (AN), ShuffleNet (SfN), and MobileNet-v2 (MN), and fine-tune them on a dataset of KCN and normal cases, each including four topographic maps. We also consider a PI classifier. Then, our EDTL method combines the output probabilities of each of the five classifiers to obtain a decision based on the fusion of probabilities. Individually, the classifier based on PI achieved 93.1% accuracy, whereas the deep classifiers reached classification accuracies over 90% only in isolated cases. Overall, the average accuracy of the deep networks over the four corneal maps ranged from 86% (SfN) to 89.9% (AN). The classifier ensemble increased the accuracy of the deep classifiers based on corneal maps to values ranging (92.2% to 93.1%) for SqN and (93.1% to 94.8%) for AN. Including in the ensemble-specific combinations of corneal maps’ classifiers and PI increased the accuracy to 98.3%. Moreover, visualization of first learner filters in the networks and Grad-CAMs confirmed that the networks had learned relevant clinical features. This study shows the potential of creating ensembles of deep classifiers fine-tuned with a transfer learning strategy as it resulted in an improved accuracy while showing learnable filters and Grad-CAMs that agree with clinical knowledge. This is a step further towards the potential clinical deployment of an improved computer-assisted diagnosis system for KCN detection to help ophthalmologists to confirm the clinical decision and to perform fast and accurate KCN treatment.

Multiclass wound image classification using an ensemble deep CNN-based classifier

Acute and chronic wounds are a challenge to healthcare systems around the world and affect many people's lives annually. Wound classification is a key step in wound diagnosis that would help clinicians to identify an optimal treatment procedure. Hence, having a high-performance classifier assists wound specialists to classify wound types with less financial and time costs. Different wound classification methods based on machine learning and deep learning have been proposed in the literature. In this study, we have developed an ensemble Deep Convolutional Neural Network-based classifier to categorize wound images into multiple classes including surgical, diabetic, and venous ulcers. The output classification scores of two classifiers (namely, patch-wise and image-wise) are fed into a Multilayer Perceptron to provide a superior classification performance. A 5-fold cross-validation approach is used to evaluate the proposed method. We obtained maximum and average classification accuracy values of 96.4% and 94.28% for binary and 91.9% and 87.7% for 3-class classification problems. The proposed classifier was compared with some common deep classifiers and showed significantly higher accuracy metrics. We also tested the proposed method on the Medetec wound image dataset, and the accuracy values of 91.2% and 82.9% were obtained for binary and 3-class classifications. The results show that our proposed method can be used effectively as a decision support system in classification of wound images or other related clinical applications.

Self-supervised learning for Environmental Sound Classification

Environmental Sound Classification (ESC) is one of the most challenging tasks in signal processing, digital forensic and machine learning. Numerous methods have been proposed to perform ESC. The conventional models’ training depends on an enormous amount of annotated data, specifically while training the deep models. This paper presents a self-supervised learning (SSL)-based deep classifier for ESC, which is an under-explored method in the field of ESC. SSL mechanism directs the model to effectively learn prototypical features from the data itself by solving a pretext task. The model proposed in this paper takes spectrogram images as input. A pretext or an auxiliary task is defined as identification of the type of data augmentation applied to the signal. The model learned by solving the pretext task is further fine-tuned for developing the deep model for ESC. The model’s performance is evaluated on two benchmark sound classification datasets, i.e. ESC-10 and DCASE 2019 Task-1(A) datasets. The experiments and results show that the SSL model attains an improvement of 12.59% and 11.17% in accuracy compared to the baseline models of the DCASE 2019 Task-1(A) and ESC-10 datasets respectively. Moreover, the model also shows competitive performance to state-of-the-art methods.

Adversarial Turing Patterns from Cellular Automata

State-of-the-art deep classifiers are intriguingly vulnerable to universal adversarial perturbations: single disturbances of small magnitude that lead to misclassification of most inputs. This phenomena may potentially result in a serious security problem. Despite the extensive research in this area, there is a lack of theoretical understanding of the structure of these perturbations. In image domain, there is a certain visual similarity between patterns, that represent these perturbations, and classical Turing patterns, which appear as a solution of non-linear partial differential equations and are underlying concept of many processes in nature. In this paper, we provide a theoretical bridge between these two different theories, by mapping a simplified algorithm for crafting universal perturbations to (inhomogeneous) cellular automata, the latter is known to generate Turing patterns. Furthermore, we propose to use Turing patterns, generated by cellular automata, as universal perturbations, and experimentally show that they significantly degrade the performance of deep learning models. We found this method to be a fast and efficient way to create a data-agnostic quasi-imperceptible perturbation in the black-box scenario. The source code is available at https://github.com/NurislamT/advTuring.

Deep Learning Based Real Age and Gender Estimation from Unconstrained Face Image towards Smart Store Customer Relationship Management

The COVID-19 pandemic markedly changed the human shopping nature, necessitating a contactless shopping system to curb the spread of the contagious disease efficiently. Consequently, a customer opts for a store where it is possible to avoid physical contacts and shorten the shopping process with extended services such as personalized product recommendations. Automatic age and gender estimation of a customer in a smart store strongly benefit the consumer by providing personalized advertisement and product recommendation; similarly, it aids the smart store proprietor to promote sales and develop an inventory perpetually for the future retail. In our paper, we propose a deep learning-founded enterprise solution for smart store customer relationship management (CRM), which allows us to predict the age and gender from a customer’s face image taken in an unconstrained environment to facilitate the smart store’s extended services, as it is expected for a modern venture. For the age estimation problem, we mitigate the data sparsity problem of the large public IMDB-WIKI dataset by image enhancement from another dataset and perform data augmentation as required. We handle our classification tasks utilizing an empirically leading pre-trained convolutional neural network (CNN), the VGG-16 network, and incorporate batch normalization. Especially, the age estimation task is posed as a deep classification problem followed by a multinomial logistic regression first-moment refinement. We validate our system for two standard benchmarks, one for each task, and demonstrate state-of-the-art performance for both real age and gender estimation.

Automatic detection of cyberbullying using multi-feature based artificial intelligence with deep decision tree classification

Recent studies have shown that cyberbullying is a rising youth epidemic. In this paper, we develop a novel automated classification model that identifies the cyberbullying texts without fitting them into large dimensional space. On the other hand, a classifier .cannot provide a limited convergent solution due to its overfitting problem. Considering such limitations, we developed a text classification engine that initially pre-processes the tweets, eliminates noise and other background information, extracts the selected features and classifies without data overfitting. The study develops a novel Deep Decision Tree classifier that utilizes the hidden layers of Deep Neural Network (DNN) as its tree node to process the input elements. The validation confirms the accuracy of classification using the novel Deep classifier with its improved text classification accuracy.

Automated post-processing of 3D-printed parts: artificial powdering for deep classification and localisation

ABSTRACT With the rapid rise in popularity in additive manufacturing, 3D printing technologies are increasingly competitive. However, current post-processing solutions are unable to cope with the varieties of printed parts, where post-production tasks differs for each part based on the desired outcome. Therefore, current post-processing treatments rely heavily on manual labor. Thus, to move towards end-to-end automated post-processing that can cater to individual processes for different parts, a fully automated vision pipeline for classifying and locating parts printed by polymer or metal powder-based processes, such as Selective Laser Sintering (SLS), binder jetting and HP Multi Jet Fusion (MJF) is proposed. The main contributions of this paper are the simulation of powder distribution on the surface of the 3D-printed parts after powder-based printing, and the incorporation of these powdered models in a sim-to-real deep learning pipeline. Our network that was trained on powdered models obtained classification and localisation results comparable to a network trained on real images. Also, the superiority of artificial powder models was shown when compared with using the original, unpowdered CAD models, especially for parts that are significantly different from their CAD models due to powder accumulation.

Identification of Autism Spectrum Disorder using Deep Neural Network

One of the acute neuro developmental disorders throughout the world today is the Autism Spectrum disorder (ASD). It is lifelong disorder which affects the behaviour and communication skill of an individual. According to world health organization 2019 report, the number of individuals diagnosed with ASD is increasing creating a threat as it is analogous to significant health care cost. Early recognition can considerably reduce the effect. In order to get rid of the time consuming and expensive diagnosis procedures for ASD, a mobile based ASD screening tool known as ASDTest app was developed. The app recorded over 1400 number of instances covering toddler, child, adolescent and adult. It is available publicly in Kaggle and UCI Machine Learning repository for research purpose. The paper gives a new approach for identification of ASD using a deep classifier. The ASD identification works in the following steps. Feature analysis explains ASD traits thereby improving the efficiency of screening process. Further, Machine Learning (ML) classifier models report ASD class type with evaluation parameters. In this analysis, an attempt is made for the incorporation of Principal Component Analysis (PCA) for feature dimension reduction followed by the usage of Deep Neural Network (DNN) for classification of ASD class type. The data upon which the techniques are applied are collected from Kaggle and UCI ML repository. The experiment result indicates that, PCA in combination with DNN provide clinically acceptable output for effective ASD identification.

Feasibility multi-center study of artificial intelligence assistance in cervical fluid-based cytology diagnosis

Objective: To propose a method of cervical cytology screening based on deep convolutional neural network and compare it with the diagnosis of cytologists. Method: The deep segmentation network was used to extract 618 333 regions of interest (ROI) from 5, 516 cytological pathological images. Combined with the experience of physicians, the deep classification network with the ability to analyze ROI was trained. The classification results were used to construct features, and the decision model was used to complete the classification of cytopathological images. Results: The sensitivity and specificity were 89.72%, 58.48%, 33.95% and 95.94% respectively. Among the smears derived from four different preparation methods, this algorithm had the best effect on natural fallout with a sensitivity of 91.10%, specificity of 69.32%, positive predictive rate of 41.41%, and negative predictive rate of 97.03%. Conclusion: Deep convolutional neural network image recognition technology can be applied to cervical cytology screening.

Psychotic Motivation for Improving Student Performance Based On Pattern Learner Features Using Deep Neural Classifier for Bipolar Disorder Students

Psychotic Motivation for Improving Student Performance Based On Pattern Learner Features Using Deep Neural Classifier for Bipolar Disorder Students

Protect privacy of deep classification networks by exploiting their generative power

Research showed that deep learning models are vulnerable to membership inference attacks, which aim to determine if an example is in the training set of the model. We propose a new framework to defend against this sort of attack. Our key insight is that if we retrain the original classifier with a new dataset that is independent of the original training set while their elements are sampled from the same distribution, the retrained classifier will leak no information that cannot be inferred from the distribution about the original training set. Our framework consists of three phases. First, we transferred the original classifier to a Joint Energy-based Model (JEM) to exploit the model’s implicit generative power. Then, we sampled from the JEM to create a new dataset. Finally, we used the new dataset to retrain or fine-tune the original classifier. We empirically studied different transfer learning schemes for the JEM and fine-tuning/retraining strategies for the classifier against shadow-model attacks. Our evaluation shows that our framework can suppress the attacker’s membership advantage to a negligible level while keeping the classifier’s accuracy acceptable. We compared it with other state-of-the-art defenses considering adaptive attackers and showed our defense is effective even under the worst-case scenario. Besides, we also found that combining other defenses with our framework often achieves better robustness. Our code will be made available at https://github.com/ChenJiyu/meminf-defense.git.

Deep Collocative Learning for Immunofixation Electrophoresis Image Analysis.

Immunofixation Electrophoresis (IFE) analysis is of great importance to the diagnosis of Multiple Myeloma, which is among the top-9 cancer killers in the United States, but has rarely been studied in the context of deep learning. Two possible reasons are: 1) the recognition of IFE patterns is dependent on the co-location of bands that forms a binary relation, different from the unary relation (visual features to label) that deep learning is good at modeling; 2) deep classification models may perform with high accuracy for IFE recognition but is not able to provide firm evidence (where the co-location patterns are) for its predictions, rendering difficulty for technicians to validate the results. We propose to address these issues with collocative learning, in which a collocative tensor has been constructed to transform the binary relations into unary relations that are compatible with conventional deep networks, and a location-label-free method that utilizes the Grad-CAM saliency map for evidence backtracking has been proposed for accurate localization. In addition, we have proposed Coached Attention Gates that can regulate the inference of the learning to be more consistent with human logic and thus support the evidence backtracking. The experimental results show that the proposed method has obtained a performance gain over its base model ResNet18 by 741.30% in IoU and also outperformed popular deep networks of DenseNet, CBAM, and Inception-v3.

Three-Dimensional Deep Noninvasive Radiomics for the Prediction of Disease Control in Patients With Metastatic Urothelial Carcinoma treated With Immunotherapy.

Immunotherapy is effective in a small percentage of patients with cancer and no reliable predictive biomarkers are currently available. Artificial Intelligence algorithms may automatically quantify radiologic characteristics associated with disease response to medical treatments. We investigated an innovative approach based on a 3-dimensional (3D) deep radiomics pipeline to classify visual features of chest-abdomen computed tomography (CT) scans with the aim of distinguishing disease control from progressive disease to immune checkpoint inhibitors (ICIs). Forty-two consecutive patients with metastatic urothelial cancer had progressed on first-line platinum-based chemotherapy and had baseline CT scans at immunotherapy initiation. The 3D-pipeline included self-learned visual features and a deep self-attention mechanism. According to the outcome to the ICIs, a 3D deep classifier semiautomatically categorized the most discriminative region of interest on the CT scans. With a median follow-up of 13.3 months (95% CI, 11.1-15.6), the median overall survival was 8.5 months (95% CI, 3.1-13.8). According to disease response to immunotherapy, the median overall survival was 3.6 months (95% CI, 2.0-5.2) for patients with progressive disease; it was not yet reached for those with disease control. The predictive accuracy of the 3D-pipeline was 82.5% (sensitivity 96%; specificity, 60%). The addition of baseline clinical factors increased the accuracy to 92.5% by improving specificity to 87%; the accuracy of other architectures ranged from 72.5% to 90%. Artificial Intelligence by 3D deep radiomics is a potential noninvasive biomarker for the prediction of disease control to ICIs in metastatic urothelial cancer and deserves validation in larger series.

A time series classification approach to non-destructive hardness testing using magnetic Barkhausen noise emission

The process setup of manufacturing processes is generally knowledge-based and carried out once for a material batch. Industry experts observe fluctuations in product quality and tool life, albeit the process setup remains unchanged. These fluctuations are mainly attributed to fluctuations in material parameters. An in-situ detection of changes in material parameters would enable manufacturers to adapt process parameters like forces or lubrication before turbulences like unexpectedly high tool wear or degradation in product quality occurs. This contribution shows the applicability of a deep learning time series classification architecture that does not rely on handcrafted feature engineering for the classification of hardness fluctuations in a sheet-metal coil using magnetic Barkhausen noise emission. This methodology is not limited to the detection of hardness fluctuations in sheet-metal coils and can potentially be applied for the in-situ material property classification in different manufacturing processes and for different material parameters.