Big Data Image Classification Based on Distributed Deep Representation Learning Model

Zhu, Meiling; Chen, Qinghua

doi:10.1109/access.2020.3011127

Cited by 16 publications

(5 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The exploration into defending CNN-LSTM models against adversarial attacks, especially within the nuanced context of Power Quality Disturbance (PQD) classification, highlights a critical area of vulnerability in the application of deep learning to essential infrastructure [14]. The traditional defense mechanisms-while innovative and effective to various extents across different domains-manifest inherent limitations when confronted with the dynamic and sophisticated nature of adversarial threats targeting the PQD classification.Adversarial training, for example, though a foundational defense mechanism, relies on a predefined set of adversarial examples, which might not encompass the full spectrum of potential attacks, particularly those that are novel or highly sophisticated.…”

Section: Dhillon Et Al Advocate For Stochastic Activation Pruningmentioning

confidence: 99%

Enhancing Model Robustness and Accuracy Against Adversarial Attacks via Adversarial Input Training

Ingle,

Pawale

2024

IJACSA

View full text Add to dashboard Cite

Adversarial attacks present a formidable challenge to the integrity of Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) models, particularly in the domain of power quality disturbance (PQD) classification, necessitating the development of effective defense mechanisms. These attacks, characterized by their subtlety, can significantly degrade the performance of models critical for maintaining power system stability and efficiency. This study introduces the concept of adversarial attacks on CNN-LSTM models and emphasizes the critical need for robust defenses.We propose Input Adversarial Training (IAT) as a novel defense strategy aimed at enhancing the resilience of CNN-LSTM models. IAT involves training models on a blend of clean and adversarially perturbed inputs, intending to improve their robustness. The effectiveness of IAT is assessed through a series of comparisons with established defense mechanisms, employing metrics such as accuracy, precision, recall, and F1-score on both unperturbed and adversarially modified datasets.The results are compelling: models defended with IAT exhibit remarkable improvements in robustness against adversarial attacks. Specifically, IAT-enhanced models demonstrated an increase in accuracy on adversarially perturbed data to 85%, a precision improvement to 86%, a recall rise to 85%, and an F1score enhancement to 85.5%. These figures significantly surpass those achieved by models utilizing standard adversarial training (75% accuracy) and defensive distillation (70% accuracy), showcasing IAT's superior capacity to maintain model accuracy under adversarial conditions.In conclusion, IAT stands out as an effective defense mechanism, significantly bolstering the resilience of CNN-LSTM models against adversarial perturbations. This research not only sheds light on the vulnerabilities of these models to adversarial attacks but also establishes IAT as a benchmark in defense strategy development, promising enhanced security and reliability for PQD classification and related applications.

show abstract

Section: Dhillon Et Al Advocate For Stochastic Activation Pruningmentioning

confidence: 99%

Enhancing Model Robustness and Accuracy Against Adversarial Attacks via Adversarial Input Training

Ingle,

Pawale

2024

IJACSA

View full text Add to dashboard Cite

show abstract

“…Zhu and Chen performed the performance test of the model they developed based on the densities, on the CIFAR-10 dataset [17]. Krishna and Kalluri performed image classification using pre-trained architectures [18].…”

Section: Related Workmentioning

confidence: 99%

Measuring the Effect of Data Augmentation in a CNN-Based Deep Neural Network Model

ÇETİNER

Metlek

2023

ICSIS

View full text Add to dashboard Cite

Traditional classification methods have difficulty in meeting the changing needs according tothe ever-increasing data piles. With the development of processors with high performance as memory andprocessing capabilities, deep learning-based methods have been widely used. A large amount of data isneeded to train a deep learning-based model, which is a computational science field. CIFAR-10, whichcontains images of 10 different objects in the world, is a benchmark dataset used effectively in imageidentification and classification. The proposed deep learning-based models should be tested in a computerenvironment in order to be used in real life. The proposed model performs the testing process with imagesthat it has never encountered during the training phase. In this article, a deep learning model is proposedthat performs classification on the CIFAR-10 dataset, which contains images of objects in the world. Aneffective classification method has been developed by removing the overfitting effect, if any, on theproposed model. Proposed model, classification process was carried out both with and without dataaugmentation. The data set used was expanded with random crop, scale transformation, vertical andhorizontal flipping data augmentation techniques. In the experimental studies, there was a big differencebetween the performance of the process using the data augmentation technique and the process without anyaugmentation. Using different augmentation techniques together or individually did not improve modelperformance. Proposed model achieved success rates of 91.93%, 93.63% and 90.49%, respectively,including train accuracy, precision, recall. According to the results obtained, it can be said that the studyhas achieved results that can compete with the literature.

show abstract

“…Recent machine learning developments and optimization allowed new image classification techniques, such as random forest and support vector machine (SVM) to be applied to information recognition (Zhu and Chen, 2020). Remote sensing-based seismic damage recognition for buildings is an image classification according to damage levels based on various characteristics.…”

Section: Introductionmentioning

confidence: 99%

Earthquake-induced building damage recognition from unmanned aerial vehicle remote sensing using scale-invariant feature transform characteristics and support vector machine classification

et al. 2023

View full text Add to dashboard Cite

Building damage is the main cause of casualties and economic losses from earthquakes. Therefore, understanding building damage is critical for emergency handling. Current information acquisition methods for assessing earthquake damage using unmanned aerial vehicle (UAV) remote sensing systems offer great flexibility and high efficiency with the capability to obtain high-resolution images, which can reflect actual damage to affected areas intuitively. Consequently, UAV remote sensing has become a convenient and important means to acquire earthquake-induced building damage information. Although manual visual interpretation can achieve high recognition accuracy, it is extremely time-consuming. In contrast, although current automatic recognition methods require less time, they have relatively poor recognition accuracy. Neither approach can satisfactorily simultaneously meet efficiency and accuracy requirements for earthquake emergency handling. This article applies image classification algorithms based on the support vector machine (SVM) to earthquake-induced building damage recognition, and proposes a recognition method based on scale-invariant feature transform (SIFT) characteristics and SVM classification. We use the magnitude 6.4 earthquake at Yangbi (2021) as an example to validate the proposed method. Results verify that the proposed method can recognize earthquake damage quickly and accurately, providing effective support for decision-making regarding rescue actions.

show abstract

Big Data Image Classification Based on Distributed Deep Representation Learning Model

Cited by 16 publications

References 32 publications

Enhancing Model Robustness and Accuracy Against Adversarial Attacks via Adversarial Input Training

Enhancing Model Robustness and Accuracy Against Adversarial Attacks via Adversarial Input Training

Measuring the Effect of Data Augmentation in a CNN-Based Deep Neural Network Model

Earthquake-induced building damage recognition from unmanned aerial vehicle remote sensing using scale-invariant feature transform characteristics and support vector machine classification

Contact Info

Product

Resources

About