Cross-entropy Research Articles

Engineering the Neural Collapse Geometry of Supervised-Contrastive Loss (Student Abstract)

Supervised-contrastive loss (SCL) is an alternative to cross-entropy (CE) for classification tasks that makes use of similarities in the embedding space to allow for richer representations. Previous works have used trainable prototypes to help improve test accuracy of SCL when training under imbalance. In this work, we propose the use of fixed prototypes to help engineering the feature geometry when training with SCL. We gain further insights by considering a limiting scenario where the number of prototypes far outnumber the original batch size. Through this, we establish a connection to CE loss with a fixed classifier and normalized embeddings. We validate our findings by conducting a series of experiments with deep neural networks on benchmark vision datasets.

ConsistentEE: A Consistent and Hardness-Guided Early Exiting Method for Accelerating Language Models Inference

Early Exiting is one of the most popular methods to achieve efficient inference. Current early exiting methods adopt the (weighted) sum of the cross entropy loss of all internal classifiers as the objective function during training, imposing all these classifiers to predict all instances correctly. However, during inference, as long as one internal classifier predicts an instance correctly, it can accelerate without losing accuracy. Thus, there is a notable gap between training and inference. We propose ConsistentEE, an early exiting method that is consistent in training and inference. ConsistentEE formulates the early exiting process as a reinforcement learning problem. A policy network is added to decide whether an instance should exit or continue. The training objective of ConsistentEE only requires each instance to be predicted correctly by one internal classifier. Additionally, we introduce the concept "Memorized Layer" to measure the hardness of an instance. We incorporate the memorized layer into reward function design, which allows "easy'' instances to focus more on acceleration while ``hard'' instances to focus more on accuracy. Experimental results show that our method outperforms other baselines on various natural language understanding and generation tasks using PLMs and LLMs as backbones respectively.

A large-scale microblog dataset and stock movement prediction based on Supervised Contrastive Learning model

The integration of Deep Neural Networks (DNN) with Natural Language Processing (NLP) technologies has opened new avenues in financial market prediction, particularly through the utilization of textual information. This study represents a significant advancement, which offers two primary contributions to stock trend prediction: (i) the exploitation of textual data (news, comments, microblogs) using advanced DNN architectures, enhancing market information utilization; (ii) significant improvement of the accuracy of predicting the direction of stock volatility by integrating textual and neural network technologies. Meanwhile, we have crawled, filtered, and constructed a large-scale microblog dataset. This dataset includes approximately 114,992 microblog textual data from 40 Science and Technology Innovation Board (STIB) companies in China during 2021. We conducted a comprehensive analysis using various DNN techniques, including Feedback Neural Networks (FNN), Supervised Contrastive Learning (SCL), Cross Entropy (CE), and Dual Contrastive Learning (DualCL), in conjunction with bag of words models, BERT, and Roberta compilers. Our findings reveal that the SCL method, when combined with microblog data, significantly increases prediction accuracy, particularly during the COVID-19 period. Furthermore, we discovered that using a cross-stock dataset enhances the accuracy of all prediction methods, and random allocation of microblog data leads to better results than sequential allocation. Additionally, we compared the efficacy of traditional models like the CAPM, three-factor, and five-factor models against neural network-based methods. Our results suggest a notable superiority of the SCL method in increasing prediction accuracy. Finally, applying our findings to real-world trading strategies, we demonstrated the practical advantages of using the SCL method in trading, evidenced by significant improvements across all performance indicators.

Rice disease segmentation method based on CBAM-CARAFE-DeepLabv3+

Rice is an important food crop, but it is susceptible to diseases during its growth process. Rapid, accurate, and effective identification of rice diseases is important for targeted measures to control disease spread. It is crucial for improving rice yield and quality. Therefore, this study proposes a CBAM-CARAFE-DeepLabv3+ rice disease segmentation method that combines attention mechanisms and feature recombination. This method focuses on three common diseases in rice growth: bacterial blight, blast, and brown spot disease. To enhance the extraction of favorable features, the algorithm adopts CBAM-RepViT as the backbone network. That is, the Squeeze-and-Excitation (SE) attention mechanism embedded in the efficient and lightweight RepViT network is replaced by the lightweight Convolutional Block Attention Module (CBAM). Compared to the SE attention mechanism, CBAM introduces a spatial attention module that focuses on important spatial positions in the feature map. It allows the model to extract more rich and detailed feature information by attending to both the channel and spatial dimensions of the feature map. Additionally, to further improve the feature extraction ability and image edge segmentation accuracy during upsampling, the lightweight Content-Aware ReAssembly of FEatures (CARAFE) operator is introduced into the decoding module for upsampling. Finally, to address the issue of imbalance between foreground and background pixel ratios in rice disease, a hybrid loss function composed of cross-entropy (CE) loss and Dice loss is proposed. Experimental results show that, compared to other networks such as DeepLabv3+, the proposed CBAM-CARAFE-DeepLabv3+ method achieves further improvement in segmentation accuracy, providing a new method for the development of rice disease segmentation technology.

Meta model-based and cross entropy-based importance sampling algorithms for efficiently solving system failure probability function

The multi-mode system failure probability function (SFPF) can quantify how the distribution parameters of the random input vector affect the system safety and decouple the system reliability-based design optimization model. However, for a problem with a time-consuming implicit performance function and rare failure domain, efficiently solving the SFPF remains significantly challenging. Therefore, in this study, two efficient algorithms are proposed, namely, the meta model-based importance sampling and cross entropy-based importance sampling. The contributions of this study are twofold. The first is constructing a single-loop optimal importance sampling density (SL-OISD) method to decouple the double-loop framework for analyzing the SFPF. The second is establishing two methods to efficiently approximate the SL-OISD and complete the SFPF estimation. The first method is based on the meta model of the system performance function, which is abbreviated as SL-Meta-IS. The second method is based on minimizing the cross entropy between the Gaussian mixture density model and SL-OISD, which is abbreviated as SL-CE-IS. To reduce the number of evaluating the system performance function when approximating the SL-OISD, sampling the SL-OISD, and identifying the state of the samples for completing the SFPF estimation, an adaptive Kriging model of the system performance function is introduced into SL-Meta-IS and SL-CE-IS. Owing to decoupling the double-loop framework into a single-loop framework, replacing the time-consuming system performance function with the economic Kriging model, and employing importance sampling variance reduction techniques to address issues related to the rare failure domain, the proposed SL-Meta-IS and SL-CE-IS methods greatly enhance the efficiency of SFPF estimations. The numerical and practical examples demonstrate that the two proposed methods are superior to the existing algorithms; moreover, the efficiency of SL-CE-IS is higher than that of SL-Meta-IS.

Image Segmentation Based On U-Net and Adjusted U-Nets

Image segmentation, the process of dividing an image into various areas is quite crucial recently. Numerous industries, including robotics, remote sensing, and medical imaging, have applications for this task. In recent years, deep learning techniques, especially U-shaped networks (U-nets), have shown remarkable success in solving image segmentation problems. This paper provides an overview of image segmentation using neural networks, introduces different types of adjusted U-nets used for this task, including the implementation of attention gates and the use of residual neural network as the encoder path based on the original encoder-decoder structure, and then uses U-net and adjusted U-nets to conduct image segmentation on the black sea sprat. The study uses dice similarity coefficient and binary cross entropy loss function to compare the model training results and further judges the functionality of the models by the predicted segmented images. According to the test results, the Res34-UNet with attention gates performs most efficiently in segmenting this image dataset, although it's more unstable compared to the basic U-Net.

Identification of rice disease under complex background based on PSOC-DRCNet

Rice is a crucial agricultural crop, yet it frequently suffers from various diseases, leading to decreased yields and, in severe cases, crop failure. Diseases significantly affect rice growth and yield, resulting in economic losses and food security challenges. The role of image recognition in identifying rice diseases is critical in agricultural production. It enables automated and efficient detection of rice diseases, which is essential for effective management, ensuring food security and sustainable agriculture. To address issues like background noise and edge blurring in rice disease image capture, as well as challenges in determining the optimal learning rate during the training of traditional rice disease recognition networks, a novel method based on PSOC-DRCNet is proposed for rice disease recognition.. First, tto solve the problem of background interference, Dual Mode Attention (DMA) is proposed to adaptively capture meaningful regions in rice disease images. Second, the Residual Adaptive Block(RAB) is proposed, which utilizes dimensional changes and channel attention to solve edge blur problems. Then, a Cross entropy and regularized mixed Loss function (CerLoss), is proposed to optimize the learning strategy of the model in the process of processing datasets and enhance the performance and generalization ability of the model to avoid overfitting problems. Ultimately, In response to the cumbersome problem of finding the optimal learning rate, we propose using Particle Swarm Optimization Chameleon (PSOC) to find the optimal learning rate and train the PSOC-DRCNet model on our custom dataset and compare it with other existing methods and the final average classification accuracy of PSOC-DRCNet is 93.88% with an F1 score of 0.940. We compare it with other existing methods. It is proved that the average classification accuracy of our model under hyper-parameter unification is 92.65% F1 score is 0.928. We validated the PSOC-DRCNet by conducting comparative analyses with other models and through generalization experiments and module effectiveness tests. Additionally, the practicality of PSOC-DRCNet was confirmed through its application in real-world scenarios. The methods proposed in this paper successfully enable the identification of various diseases in rice leaves, offering a practical solution for incorporating deep learning into the agricultural production process. Furthermore, these findings serve as a valuable reference for disease identification in other crops.

SCT: Summary Caption Technique for Retrieving Relevant Images in Alignment with Multimodal Abstractive Summary

This work proposes an efficient Summary Caption Technique that considers the multimodal summary and image captions as input to retrieve the correspondence images from the captions that are highly influential to the multimodal summary. Matching a multimodal summary with an appropriate image is a challenging task in computer vision and natural language processing. Merging in these fields is tedious, though the research community has steadily focused on cross-modal retrieval. These issues include the visual question-answering, matching queries with the images, and semantic relationship matching between two modalities for retrieving the corresponding image. Relevant works consider questions to match the relationship of visual information and object detection and to match the text with visual information and employing structural-level representation to align the images with the text. However, these techniques are primarily focused on retrieving the images to text or for image captioning. But less effort has been spent on retrieving relevant images for the multimodal summary. Hence, our proposed technique extracts and merge features in the Hybrid Image Text layer and captions in the semantic embeddings with word2vec where the contextual features and semantic relationships are compared and matched with each vector between the modalities, with cosine semantic similarity. In cross-modal retrieval, we achieve top five related images and align the relevant images to the multimodal summary that achieves the highest cosine score among the retrieved images. The model has been trained with seq-to-seq modal with 100 epochs, besides reducing the information loss by the sparse categorical cross entropy. Further, experimenting with the multimodal summarization with multimodal output dataset, in cross-modal retrieval, helps to evaluate the quality of image alignment with an image-precision metric that demonstrate the best results.

Semantic segmentation of thermal defects in belt conveyor idlers using thermal image augmentation and U-Net-based convolutional neural networks

The belt conveyor (BC) is the main means of horizontal transportation of bulk materials at mining sites. The sudden fault in BC modules may cause unexpected stops in production lines. With the increasing number of applications of inspection mobile robots in condition monitoring (CM) of industrial infrastructure in hazardous environments, in this article we introduce an image processing pipeline for automatic segmentation of thermal defects in thermal images captured from BC idlers using a mobile robot. This study follows the fact that CM of idler temperature is an important task for preventing sudden breakdowns in BC system networks. We compared the performance of three different types of U-Net-based convolutional neural network architectures for the identification of thermal anomalies using a small number of hand-labeled thermal images. Experiments on the test data set showed that the attention residual U-Net with binary cross entropy as the loss function handled the semantic segmentation problem better than our previous research and other studied U-Net variations.

MRI-CE: Minimal rare itemset discovery using the cross-entropy method

Rare itemsets have been studied less extensively than frequent itemsets, but have important potential applications in black swan events, like detecting anomalies. Mining rare itemsets poses two challenges: too many results may be obtained, and the process may incur a high computational overhead. To overcome these two challenges, we can attempt to mine minimal rare itemsets (MRIs) and use heuristic methods to mine approximate results instead of exact results. This paper describes a novel algorithm for mining MRIs using cross-entropy (CE). We present the modeling method for MRI-CE and introduce a progressive checking strategy that enables more MRIs to be discovered in each iteration. The discovered MRIs are then used to update a probability vector. We design two optimization strategies to improve the algorithm's performance. The adaptive sample size strategy narrows the search space as the number of iterations increases, and the crossover-based individual generation strategy improves the diversity of the samples. To evaluate the performance of MRI-CE, we select six competitive algorithms and conduct extensive experiments on publicly available datasets. The results show that the proposed algorithm is not only efficient, but also highly accurate. Furthermore, we verify the effectiveness of the two optimization strategies experimentally.

Interval-based sparse ensemble multi-class classification algorithm for terahertz data

Terahertz time-domain spectroscopy (THz-TDS) has been widely used for food and drug identification. The classification information of a THz spectrum usually does not exist in the whole spectral band but exists only in one or several small intervals. Therefore, feature selection is indispensable in THz-based substance identification. However, most THz-based identification methods empirically intercept the low-frequency band of the THz absorption coefficients for analysis. In order to adaptively find out important intervals of the THz spectra, an interval-based sparse ensemble multi-class classifier (ISEMCC) for THz spectral data classification is proposed. In ISEMCC, the THz spectra are first divided into several small intervals through window sliding. Then the data of training samples in each interval are extracted to train some base classifiers. Finally, a final robust classifier is obtained through a nonnegative sparse combination of these trained base classifiers. With l1 -norm, two objective functions that based on Mean Square Error (MSE) and Cross Entropy (CE) are established. For these two objective functions, two iterative algorithms based on the Alternating Direction Method of Multipliers (ADMM) and Gradient Descent (GD) are built respectively. ISEMCC transforms the problem of interval feature selection and decision-level fusion into a nonnegative sparse optimization problem. The sparse constraint ensures only a few important spectral segments are selected. In order to verify the performance of the proposed algorithm, comparative experiments on identifying the origin of Bupleurum and the harvesting year of Tangerine peel are carried out. The base classifiers used by ISEMCC are Support Vector Machine (SVM) and Decision Tree (DT). The experimental results demonstrate that the proposed algorithm outperforms six typical classifiers, including Random Forest (RF), AdaBoost, RUSBoost, ExtraTree, and the two base classifiers, in terms of classification accuracy.

A joint autoencoder and classifier deep neural network for AD and MCI classification

AbstractIn this article, we present a new approach to distinguish progressive mild cognitively impaired (pMCI) subjects, who eventually develop Alzheimer's disease (AD) from stable MCI (sMCI) subjects whose situation does not deteriorate into AD. The proposed approach combines the discriminating capabilities of classifiers and representation learning capacities of autoencoders into a unified architecture, and is hence termed as joint autoencoder and classifier deep neural network (JACDNN). JACDNN employs a single classifier and multiple autoencoders that are trained together to perform pattern classification. The classifier in JACDNN is trained using standard approaches to distinguish between subject from different classes using the binary cross entropy loss. The autoencoders in JACDNN, regularizes individual layers in the network used for classification to learn representations useful for reconstructing a given input. The performance of JACDNN has been evaluated on several machine learning problems pertaining to dementia, namely AD versus cognitively normal (CN) subjects, AD versus sMCI, CN versus pMCI, and pMCI versus sMCI. These problems are targeted using two datasets. The first dataset consist of gray matter (GM) features of subjects and the second dataset consist of combination of GM and white matter (WM) features. It is observed that better classification results are obtained when the classifier is built on GM and WM as compared with GM features alone. Performance comparison of JACDNN with other existing approaches has been conducted for these problems. The results clearly indicate that JACDNN performs better than other existing approaches for these problems.

Segmentation method of magnetic resonance imaging brain tumor images based on improved UNet network.

Glioma is a primary malignant craniocerebral tumor commonly found in the central nervous system. According to research, preoperative diagnosis of glioma and a full understanding of its imaging features are very significant. Still, the traditional segmentation methods of image dispensation and machine wisdom are not acceptable in glioma segmentation. This analysis explores the potential of magnetic resonance imaging (MRI) brain tumor images as an effective segmentation method of glioma. This study used 200 MRI images from the affiliated hospital and applied the 2-dimensional residual block UNet (2DResUNet). Features were extracted from input images using a 2×2 kernel size (64-kernel) 1-step 2D convolution (Conv) layer. The 2DDenseUNet model implemented in this study incorporates a ResBlock mechanism within the UNet architecture, as well as a Gaussian noise layer for data augmentation at the input stage, and a pooling layer for replacing the conventional 2D convolutional layers. Finally, the performance of the proposed protocol and its effective measures in glioma segmentation were verified. The outcomes of the 5-fold cross-validation evaluation show that the proposed 2DResUNet and 2DDenseUNet structure has a high sensitivity despite the slightly lower evaluation result on the Dice score. At the same time, compared with other models used in the experiment, the DM-DA-UNet model proposed in this paper was significantly improved in various indicators, increasing the reliability of the model and providing a reference and basis for the accurate formulation of clinical treatment strategies. The method used in this study showed stronger feature extraction ability than the UNet model. In addition, our findings demonstrated that using generalized die harm and prejudiced cross entropy as loss functions in the training process effectively alleviated the class imbalance of glioma data and effectively segmented glioma. The method based on the improved UNet network has obvious advantages in the MRI brain tumor portrait segmentation procedure. The result showed that we developed a 2D residual block UNet, which can improve the incorporation of glioma segmentation into the clinical process.

Multi-attribute decision making based on VIKOR with probabilistic linguistic term sets: An application to the risk evaluation of foreign direct investment.

The multiple global environments have triggered changes in the international environment, leading to a sharp decline of foreign direct investment (FDI) compared to pre-pandemic level. To evaluate the investment risk of FDI and make optimal investment decision becomes the most important issue for investors. This paper focuses on the evaluation of investment risk for FDI. First, an index system for risk evaluation of FDI is constructed. Then, we introduce the probabilistic linguistic entropy and cross entropy measures, based on which, a programming model is developed to identify the objective attribute weights. A composite weight derivation method, which takes both the objective attribute weights and the subjective attribute weights into account, is further introduced. In view of attributes' uncertainty and fuzziness and the conflicting characteristics of some attributes, the VIKOR (the Serbian name: VlseKriterijumska Optimizacija I Kompromisno Resenje, means multi-criteria optimization and compromise solution) method is used to evaluate the risk of FDI under the probabilistic linguistic environment. Furthermore, a case study is presented to illustrate the proposed method. The comparative analysis and some further discussions verify the validity of the proposed method for the FDI risk evaluation.

STGIC: A graph and image convolution-based method for spatial transcriptomic clustering.

Spatial transcriptomic (ST) clustering employs spatial and transcription information to group spots spatially coherent and transcriptionally similar together into the same spatial domain. Graph convolution network (GCN) and graph attention network (GAT), fed with spatial coordinates derived adjacency and transcription profile derived feature matrix are often used to solve the problem. Our proposed method STGIC (spatial transcriptomic clustering with graph and image convolution) is designed for techniques with regular lattices on chips. It utilizes an adaptive graph convolution (AGC) to get high quality pseudo-labels and then resorts to dilated convolution framework (DCF) for virtual image converted from gene expression information and spatial coordinates of spots. The dilation rates and kernel sizes are set appropriately and updating of weight values in the kernels is made to be subject to the spatial distance from the position of corresponding elements to kernel centers so that feature extraction of each spot is better guided by spatial distance to neighbor spots. Self-supervision realized by Kullback-Leibler (KL) divergence, spatial continuity loss and cross entropy calculated among spots with high confidence pseudo-labels make up the training objective of DCF. STGIC attains state-of-the-art (SOTA) clustering performance on the benchmark dataset of 10x Visium human dorsolateral prefrontal cortex (DLPFC). Besides, it's capable of depicting fine structures of other tissues from other species as well as guiding the identification of marker genes. Also, STGIC is expandable to Stereo-seq data with high spatial resolution.

Segmentation of brain MRI using moth-flame optimization with modified cross entropy based fitness function

Segmentation of brain MRI using moth-flame optimization with modified cross entropy based fitness function

Balanced self-distillation for long-tailed recognition

In long-tailed recognition tasks, the knowledge distillation technology is widely adopted for improving performance of deep neural networks. These methods distill the knowledge from the pretrained teacher model to the student model, which enables higher long-tailed recognition accuracy. However, the dependence on accompanying assistive models complicates the single network’s training process in the need for large memory and time costs. In this work, we present Balanced Self-Distillation (BSD) to distill tail knowledge by a single network without the assistive models. Specifically, BSD distills knowledge between different distortions of the same samples to stimulate the representation learning potential of the single network and adopts a balanced class weight for shifting the distillation focus from head-to-tail classes. Comprehensive experimentation across diverse datasets, including CIFAR-10-LT, CIFAR-100-LT and TinyImageNet-LT, consistently outperforms robust baseline methods. Specifically, BSD achieves improvements of 8.13% on CIFAR-100-LT with an imbalance ratio of 100 compared to the baseline (cross entropy). Furthermore, the proposed method enables seamless integration with contemporary techniques like re-sampling, meta-learning, and cost-sensitive learning. It emerges as a versatile tool capable of effectively addressing the challenges of long-tailed scenarios.

An efficient cross-entropy method addressing high-dimensional dependencies for composite systems reliability evaluation

Cross-entropy (CE) based importance sampling (IS) accelerates the reliability evaluation of power system greatly, but is mainly focused on the CEIS of independent random variables (RVs) or low-dimensional correlated RVs (CRVs). To extend the CE method to high-dimensional CRVs while avoiding the “curse of dimensionality” in optimizing a high-dimensional IS probability density function (IS-PDF), an efficient CE method termed as CE-DRDM, is developed from a dimension-reduced dependence model (DRDM). First, based on the DRDM’s original hierarchical disaggregate structure (HDS), the CE-DRDM which conducts the CEIS for a single aggregate RV rather than the high-dimensional CRVs is proposed. Second, to solve the issue that the performance of CE-DRDM may degrade in the case of high penetration of renewable energy, the CE-DRDM is enhanced by improving the DRDM’s HDS, and then a novel CE optimization method is proposed, through which the analytical updating formulas of IS-PDF parameters can be derived. Thirdly, the CE-DRDM is further enhanced by using a sparse density estimator, based on which the number of IS-PDF parameters needed to be optimized in the CE optimization reduces greatly, and consequently the optimization accuracy improves accordingly. Finally, the validity of CE-DRDM is verified by several numerical cases with high-dimensional dependencies.

SME-MFP: A novel spatiotemporal neural network with multiangle initialization embedding toward multifunctional peptides prediction

As a promising alternative to conventional antibiotic drugs in the biomedical field, functional peptide has been widely used in disease treatment owing to its low toxicity, high absorption rate, and biological activity. Recently, several machine learning methods have been developed for functional peptide prediction. However, the main research heavily relies on statistical features and few consider multifunctional peptide identification. So, we propose SME-MFP, a novel predictor in the imbalanced multi-label functional peptide datasets. First, we employ physicochemical and evolutionary information to represent the peptide sequence’s initialization features from multiple perspectives. Second, the features are fused and then put into spatial feature extractors, where the residual connection and multiscale convolutional neural network extract more discriminative features of different lengths’ peptide sequences. Besides, we also design AFT-based temporal feature extractors to fully capture the global interactions of the sequences. Finally, devising a new loss to replace the traditional cross entropy loss to settle the class imbalance problems. The results show that our framework not only enhances the model's ability to capture sequence features effectively, but also accuracy improves by 3.89% over existing methods on public peptide datasets.

MULTI-LABEL CLASSIFICATION OF THORACIC DISEASES USING STRUCTURED DEEP LEARNING FRAMEWORK

The current medical scenario indicates that thoracic diseases are the primary cause of illnesses of human beings worldwide. COVID-19 outbreak results in the generation of tremendous Chest X-Ray (CXR) and Computed Tomography (CT) image data archives because CXR and CT imaging are the foremost effective screening tools that characterize patterns and features of pathologies present in the lungs. After careful observation, these chest radiographs show that images are often labeled with more than one pathology, that extends disease diagnosis problem to tedious multi-label classification task. So, this paper proposes a modified mish activation function and compound loss-based Separable Convolution Neural Network (SCNN) model that accomplishes image-level detection of multiple pathologies and COVID-19 from multi-modality chest radiographs. We investigate the power of separable convolution in the CNN network that interprets spatial and depth-wise dimensions of each pixel in CXR and CT images. A modified mish called as Swmish, a non-monotonic activation function, is introduced that preserves negative gradient flow of input and capture fine-grained detail of multiple regions. Moreover, a compound loss function of dice coefficient and binary cross entropy is applied that strongly optimizes the model during training. The generality of the proposed SCNN model is confirmed after conducting comprehensive experiments on multi-label CXR and COVID19-CT datasets. The proposed model gains an AUC score of 0.91 and loss of 0.13 on the multi-label CXR dataset and achieves a testing accuracy of 0.97, AUC score of 0.99 and a loss of 0.11 on the COVID19-CT dataset. These results show that the proposed model achieves highest evaluation metrics in classifying binary and multi-label thoracic abnormalities from multi-modality chest radiographs.