L1 Loss Research Articles

Matrix factorization with a sigmoid-like loss control

Matrix factorization is one of the fundamental approaches of recommender systems. With the popular L2 loss, learning models tend to overfit significantly deviated predictions. However, predicting the actual rating of 5 as 1 or 2 makes no essential difference in the application. In this paper, we design a sigmoid-like function to control the loss of each individual prediction, which has two advantages. First, it reduces the loss corresponding to significantly deviated predictions. Therefore, the impact of these predictions, some of which may be caused by outliers, is also reduced. Second, it is independent of two classical over-fitting control techniques using regular terms and validation data, respectively. Hence, it can be combined with them to form a more powerful method. Experiments are undertaken on six benchmark datasets in comparison with different losses. Results show that the proposed loss function has good performance in terms of MAE, RMSE, and NDCG, however not so good in terms of HR and MAP.

Research on Singularity in Calculation of Induced Electric Field in Pointed Conductors

In the electric field environment of thunderstorm, the tip of a grounded object at a high place is more likely to produce corona discharge and finally cause lightning flash. Therefore, the study of the surface electric field of the grounded object tip in the electric field environment has important theoretical value and practical significance for the analysis of the characteristics of the head-on discharge. Due to the strong singularity of the tip structure, the traditional numerical analysis method cannot accurately calculate its surface electric field. Therefore, the traditional literature usually chamfers the conductor tip, but in lightning physics, researchers are most concerned about the maximum field strength in the local area. In order to study the electric field distribution on the tip conductor surface, a semi-analytical boundary element method is proposed to calculate the tip structure directly. Firstly, the analytical formula of the electric field on the tip conductor surface and the boundary element semi-analytical common method are derived, and the tip conductor model is established to calculate the tip electric field excited by the external electric field using the analytical solution and the semi-analytical method. The correctness of the semi-analytical method is verified by calculating the L2 loss function of the two, and the influence of the tip angle on the semi-analytical calculation results is studied. The results show that for the three-dimensional cone model, the L2 loss function is 10-5, which meets the calculation accuracy. The special treatment of the semi-analytical method at the strong singular integral makes the calculation accuracy much higher than that of the Gaussian integral, where the L2 loss function is greatly reduced. Further, based on the change of tip angle, as the conductor tip becomes sharper, the greater the electric field distortion is, and the greater the L2 loss function is. The research results provide a method reference for the calculation of the electric field on the tip conductor surface.

Predicting Individual Survival Distributions Using ECG: A Deep Learning Approach Utilizing Features Extracted by a Learned Diagnostic Model

In the field of healthcare, individual survival prediction is important for personalized treatment planning. This study presents machine learning algorithms for predicting Individual Survival Distributions (ISD) using electrocardiography (ECG) data in two different formats. The models, which predict time until death, are developed and evaluated on a large, population-based cohort from Alberta, Canada. Our results demonstrate that models trained on raw ECG waveforms significantly outperform those trained on traditional ECG measurements in several metrics, including concordance index, hinge L1 loss, margin L1 loss, and margin truncated L1 loss. Additionally, the integration of predicted probabilities from wide-range diagnostic tasks not only enhances our ISD models' performance but also makes them significantly superior to other models across all evaluation metrics in individual survival prediction tasks. This innovative approach highlights the potential to leverage insights from diagnostic models for prognostic tasks, such as individual survival prediction. These findings could have far-reaching implications for the development of personalized treatment plans and open new avenues for future research in survival prediction using ECGs.

A Robust Approach to Object Detection in Images using Improved T_CenterNet Method

Aims: A Robust Approach to Object Detection in Images using Improved T_CenterNet Method Background: Currently, two-stage image object detectors are the most common method of detecting objects. However, due to the slow processing of two-stage detectors, it is not always possible to apply them in real-time. Objective: To develop a more robust and effective method for detecting objects in images by building upon and improving the T_CenterNet algorithm Method: In this research paper, we propose CenterNet, a method based on a one-stage detector that takes into consideration the heat map branch of CenterNet, which is the most accurate anchorfree detector. To improve the outcomes of the upcoming image as well as detection efficiency, we disseminate the previously accurate long-term detection using a heatmap and a cascade guiding framework. Result: On the COCO dataset, our technique obtains detection outcome AP50 -65.5%, AP! (Small Objects)-33.8, AP! (Medium Objects)-47.0 at 38 frames per second. The results of the study show the superiority of the suggested approach over the conventional one. Conclusion: In this research, we have introduced an enhanced approach to image object detection using the improved T_CenterNet method. While the original CenterNet algorithm demonstrated superior precision compared to other single-stage algorithms, it suffered from slow detection speed. Our proposed approach, the Hourglass-208 model, built upon CenterNet, significantly improved detection performance. The proposed T_CenterNet algorithm improves upon the original CenterNet by incorporating the Hourglass backbone and feature map fusion techniques, enabling enhanced object detections. Additionally, we applied the Smooth L1 loss function to object size estimation within T_CenterNet, greatly enhancing overall detection performance by improving object size regression.

Perception-Distortion Balanced Super-Resolution: A Multi-Objective Optimization Perspective.

High perceptual quality and low distortion degree are two important goals in image restoration tasks such as super-resolution (SR). Most of the existing SR methods aim to achieve these goals by minimizing the corresponding yet conflicting losses, such as the l1 loss and the adversarial loss. Unfortunately, the commonly used gradient-based optimizers, such as Adam, are hard to balance these objectives due to the opposite gradient decent directions of the contradictory losses. In this paper, we formulate the perception-distortion trade-off in SR as a multi-objective optimization problem and develop a new optimizer by integrating the gradient-free evolutionary algorithm (EA) with gradient-based Adam, where EA and Adam focus on the divergence and convergence of the optimization directions respectively. As a result, a population of optimal models with different perception-distortion preferences is obtained. We then design a fusion network to merge these models into a single stronger one for an effective perception-distortion trade-off. Experiments demonstrate that with the same backbone network, the perception-distortion balanced SR model trained by our method can achieve better perceptual quality than its competitors while attaining better reconstruction fidelity. Codes and models can be found at https://github.com/csslc/EA-Adam.

Time series analysis of GSS bonds

Abstract The market for green bonds, and environmentally aligned investment solutions, is increasing. As of 2022, the market of green bonds exceeded USD 2 trillion in issuance, with India, for example, having issued its first-ever sovereign green bonds totally R80bn (c.USD1bn) in January 2023. This paper lays the foundation for future papers and summarises the initial stages of our analysis, where we try to replicate the S&P Green Bond Index (i.e. this is a time series problem) over a period using non-traditional techniques. The models we use include neural networks such as CNNs, LSTMs and GRUs. We extend our analysis and use an open-source decision tree model called XGBoost. For the purposes of this paper, we use 1 day’s prior index information to predict today’s value and repeat this over a period of time. We ignore for example stationarity considerations and extending the input window/output horizon in our analysis, as these will be discussed in future papers. The paper explains the methodology used in our analysis, gives details of general underlying background information to the architecture models (CNNs, LSTMs, GRUs and XGBoost), as well as background to regularisation techniques specifically L2 regularisation, loss curves and hyperparameter optimisation, in particular, the open-source library Optuna.

A Study of the Intersection of Traditional Calligraphy and Digital Media Art

Abstract As a classic representative of traditional culture, Chinese calligraphy art has more and more obvious advantages in today’s digital media art creation. In this paper, the calligraphic text image is binarized by the OTSU method, the image enhancement is realized by linear gray value stretching, and the optimal tilt correction method is used to preprocess the traditional calligraphic text image and make a font library. Based on the data from the calligraphy font library, an immersive digital interactive context of traditional calligraphy is designed using digital media art. Pix2Pix network is introduced to combine with the PatchGAN network to establish a style migration network for traditional calligraphy text images, and three different types of loss functions are designed to ensure the accurate generation of traditional calligraphy text. The L1 loss of this paper’s model in generating traditional calligraphy characters is only 0.0359 at the lowest, the similarity values of the generated samples of calligraphy characters are mainly between 90% and 96%, and the maximum error of the relative position and relative size of the radicals is only 1.18%. The average value of the G2R indicator for the users’ images of calligraphic characters was 75.23%, and 89.86% of the users indicated that they were basically satisfied with the immersive interactive context of traditional calligraphy. Traditional calligraphy can achieve digital transformation and provide audiences with more diverse cultural experiences by relying on digital media art.

Research on Self-service Customs Clearance System at Border Crossings Based on Deep Learning Models

Abstract This paper proposes a deep learning method for face recognition in the self-service customs clearance system at border crossings and designs the encoder and face feature mining module in the learning framework. Meanwhile, the loss function is constructed by combining L1 loss and KL scatter. The face recognition technology based on the deep learning model is used to construct the self-service border crossing system, and the research and analysis are conducted from two aspects, namely, the test of the self-service border crossing system and the application situation. The number of outbound self-clearance acceptors has increased by 2957931, and the self-clearance system at border crossings is able to provide more travelers with the convenience brought by self-clearance. This study solves the problem of self-clearance at border crossing with the help of face recognition technology in a deep learning model, which provides technical support and theoretical reference for the optimization and upgrading of self-clearance system at border crossing in the future.

Fractal Dimension on CBCT Images and Modular Neural Networks to Identify Reduced Bone Mineral Density in Women

This paper provides two different methods to diagnose osteoporosis in women; the first method is the fractal analysis evaluated by CBCT at two bone locations (the mandible and the second cervical vertebrae) to see if there is any correlation between the two. At the same time, the second method is deep convolutional neural networks (DCNNs). One hundred eighty-eight patients' mandibular CBCT images were used, and DCNN models based on the ResNet-101 framework were employed. Dual X-ray absorptiometry of the hip and lumbar spine revealed that 139 of the 188 postmenopausal women tested had osteoporosis, whereas 49 had average bone mineral density. The second cervical vertebra and the mandible were selected as locations of interest for FD analysis on the CBCT images. Measurement accuracy, both within and between observers' agreements, and correlations between two data sets were all calculated. To evaluate osteoporosis, we used a segmented, three-phase approach. Stage 1 was devoted to the identification of mandibular bone slices. In Stage 2, the coordinates for the mandible's cross-sectional views were established, and Stage 3 calculated the thickness of the mandible bone, emphasizing osteoporotic variations. The average FD values within the interest area of the mandible were significantly lower in people with osteoporosis than in those with average bone mineral density. At the same time, the two groups had no significant difference in FD values at the second cervical vertebra. For the mandibular site, areas beneath the curve were 0.644 (P = 0.008), while the area under the curve for the vertebral site was 0.531 (P = 0.720). DCNN training in the first stage yielded an astounding 98.85% training accuracy, the second stage decreased L1 loss to a meager 1.02 pixels, and the bone thickness computation method used in the last stage had a mean squared error of 0. 8377. We concluded that FD was underutilized even though it distinguished between women with normal BMD and those with osteoporosis in the mandibular area. Additionally, even with small mandibular CBCT datasets, the results show the value of a modular transfer learning approach for osteoporosis detection.

An unconstrained palmprint region of interest extraction method based on lightweight networks.

Accurately extracting the Region of Interest (ROI) of a palm print was crucial for subsequent palm print recognition. However, under unconstrained environmental conditions, the user's palm posture and angle, as well as the background and lighting of the environment, were not controlled, making the extraction of the ROI of palm print a major challenge. In existing research methods, traditional ROI extraction methods relied on image segmentation and were difficult to apply to multiple datasets simultaneously under the aforementioned interference. However, deep learning-based methods typically did not consider the computational cost of the model and were difficult to apply to embedded devices. This article proposed a palm print ROI extraction method based on lightweight networks. Firstly, the YOLOv5-lite network was used to detect and preliminarily locate the palm, in order to eliminate most of the interference from complex backgrounds. Then, an improved UNet was used for keypoints detection. This network model reduced the number of parameters compared to the original UNet model, improved network performance, and accelerated network convergence. The output of this model combined Gaussian heatmap regression and direct regression and proposed a joint loss function based on JS loss and L2 loss for supervision. During the experiment, a mixed database consisting of 5 databases was used to meet the needs of practical applications. The results showed that the proposed method achieved an accuracy of 98.3% on the database, with an average detection time of only 28ms on the GPU, which was superior to other mainstream lightweight networks, and the model size was only 831k. In the open-set test, with a success rate of 93.4%, an average detection time of 5.95ms on the GPU, it was far ahead of the latest palm print ROI extraction algorithm and could be applied in practice.

Deep learning for high-resolution multichannel seismic impedance inversion

Seismic impedance inversion can obtain subsurface physical properties and plays an important role in hydrocarbon and mineral exploration. Due to the inaccurate and insufficient seismic data, the inverse problem is ill posed as characterized by unreliability and nonuniqueness of solutions. Regularization techniques relying on certain prior information often are introduced to force the inverse problem to obtain stable results with predetermined characteristics. However, for complex geologic conditions, these methods usually have difficulty achieving satisfactory accuracy and resolution. We develop a deep-learning (DL)-based multichannel impedance inversion method that flexibly incorporates prior information by training with numerous realistic structural 2D impedance models based on the features of field data. The DL framework is supplemented by the attention mechanism and residual block to automatically learn more features and details from training data. A novel hybrid loss function, combining [Formula: see text] loss and multiscale structural similarity loss, is introduced to enhance the network’s capacity for learning structural features. Synthetic and field data examples demonstrate that our method can effectively produce inversion results with high resolution, good lateral continuity, and enhanced structural features compared with traditional methods.

Fabric Defect Detection Method Using SA-Pix2pix Network and Transfer Learning

This paper proposes a fabric defect detection algorithm based on the SA-Pix2pix network and transfer learning to address the issue of insufficient accuracy in detecting complex pattern fabric defects in scenarios with limited sample data. Its primary contribution lies in treating defects as disruptions to the fabric’s texture. It leverages a generative adversarial network to reconstruct defective images, restoring them to images of normal fabric texture. Subsequently, the reconstituted images are subjected to dissimilarity calculations against defective images, leading to image segmentation for the purpose of defect detection. This approach addresses the issues of poor defect image reconstruction accuracy due to the limited ability of remote dependency modeling within the generator’s convolutional neural network. It also tackles deficiencies in the generative adversarial network’s loss function in handling image details. To enhance the structure and loss function of the generative adversarial network, it introduces self-attention mechanisms, L1 loss, and an improved structural loss, thus mitigating the problems of low defect image reconstruction accuracy and insufficient image detail handling by the network. To counteract the issue of declining model training accuracy in the face of sparse complex fabric defect samples, a channel-wise domain transfer learning approach is introduced. This approach constrains the training of the target network through feature distribution, thereby overcoming the problem of target network overfitting caused by limited sample data. The study employs three methods to experimentally compare and investigate five distinct complex pattern fabric defects. The results demonstrate that, when compared to two other defect detection methods, the approach advocated in this paper exhibits superior detection accuracy in scenarios with limited sample data.

A Multistage Rigid-Affine-Deformable Network for Three-Dimensional Multimodal Medical Image Registration

Multimodal image registration is an important component of medical image processing, allowing the integration of complementary information from various imaging modalities to improve clinical applications like diagnosis and treatment planning. We proposed a novel multistage neural network for three-dimensional multimodal medical image registration, which addresses the challenge of larger rigid deformations commonly present in medical images due to variations in patient positioning in different scanners and rigid anatomical structures. This multistage network combines rigid, affine and deformable transformations in three stages. The network was trained unsupervised with Mutual Information and Gradient L2 loss. We compared the results of our proposed multistage network with a rigid-affine-deformable registration with the classical registration method NiftyReg as a baseline and a multistage network, which combines affine and deformable transformation, as a benchmark. To evaluate the performance of the proposed multistage network, we used four three-dimensional multimodal in vivo datasets: three renal MR datasets consisting of T1-weighted and T2-weighted MR scans and one liver dataset containing CT and T1-weighted MR scans. Experimental results showed that combining rigid, affine and deformable transformations in a multistage network leads to registration results with a high structural similarity, overlap of the corresponding structures (Dice: 76.7 ± 12.5, 61.1 ± 14.0, 64.8 ± 16.2, 68.1 ± 24.6 for the four datasets) and a low level of image folding (|J| ≤ 0: less than or equal to 1.1%), resulting in a medical plausible registration result.

Clinical case of pulmonary cement embolism and adjacent fractures as a complication after vertebroplasty

Introduction: Osteoporosis is a skeleton disease with reduction in bone mass and disruption of the bone microstructure. A rare secondary form is pregnancy-associated osteoporosis. The standard method for determining bone density is DEXA measurement (dual X-ray absorptiometry). With the radiofrequency echographic multi spectrometry (REMS) method this is done using ultrasound without radiation exposure. Osteoporosis often leads to pathological fractures of the vertebrae. In case of severe pain and immobilization vertebroplasty is considered standard procedure. However, it can have side effects such as cement leakage into surrounding tissue, pulmonary embolism and adjacent fractures.
 Case: We present a 27-year-old female patient with compression fractures of the Th8 and Th12 5 months after labour. DEXA measurements showed values of the T-score at hip neck -2,5 SD (total) and -3,5 SD (L1-L4) at spine. The fractures of the Th8 and Th12 were then treated with vertebroplasty. A post-operative CT showed cement leakage and extravasation, as well as bilateral cement embolism of the proximal branches of the pulmonary arteries and progression of the height loss of Th5, Th7 and L1.
 Discussion: Pulmonary cement embolism is a rare complication of vertebroplasty. Cement leakage is more frequent complication, although most of the time asymptomatic. These risks of complications make the alternative conservative treatment worth considering before proceeding to surgical interventions. Only after failure to achieve adequate pain management kyphoplasty or vertebroplasty should be considered, due to the risks of prolonged immobilization of the patient leading to further bone and muscle loss.
 Conclusion: Pregnancy associated osteoporosis is very rare. The radiation free REMS method is suitable for pregnant women. Vertebroplasty offers pain relief and rapid mobilization of the patient. However, it poses a risk of numerous complications. The osteoporotic fractures of the vertebrae are usually stable and a conservative treatment prior to surgical interventions should always be considered.

MedFusionGAN: multimodal medical image fusion using an unsupervised deep generative adversarial network

PurposeThis study proposed an end-to-end unsupervised medical fusion generative adversarial network, MedFusionGAN, to fuse computed tomography (CT) and high-resolution isotropic 3D T1-Gd Magnetic resonance imaging (MRI) image sequences to generate an image with CT bone structure and MRI soft tissue contrast to improve target delineation and to reduce the radiotherapy planning time.MethodsWe used a publicly available multicenter medical dataset (GLIS-RT, 230 patients) from the Cancer Imaging Archive. To improve the models generalization, we consider different imaging protocols and patients with various brain tumor types, including metastases. The proposed MedFusionGAN consisted of one generator network and one discriminator network trained in an adversarial scenario. Content, style, and L1 losses were used for training the generator to preserve the texture and structure information of the MRI and CT images.ResultsThe MedFusionGAN successfully generates fused images with MRI soft-tissue and CT bone contrast. The results of the MedFusionGAN were quantitatively and qualitatively compared with seven traditional and eight deep learning (DL) state-of-the-art methods. Qualitatively, our method fused the source images with the highest spatial resolution without adding the image artifacts. We reported nine quantitative metrics to quantify the preservation of structural similarity, contrast, distortion level, and image edges in fused images. Our method outperformed both traditional and DL methods on six out of nine metrics. And it got the second performance rank for three and two quantitative metrics when compared with traditional and DL methods, respectively. To compare soft-tissue contrast, intensity profile along tumor and tumor contours of the fusion methods were evaluated. MedFusionGAN provides a more consistent, better intensity profile, and a better segmentation performance.ConclusionsThe proposed end-to-end unsupervised method successfully fused MRI and CT images. The fused image could improve targets and OARs delineation, which is an important aspect of radiotherapy treatment planning.

Seismic attribute analysis with a combination of convolutional autoencoder and random forest in a turbidite reservoir

Due to the complex depositional environment of a turbidite reservoir in the Niuzhuang Delta, China, the traditional seismic facies classification is a challenge to perform accurately and continuously. Due to the thin turbidite layers in the reservoir, machine-learning-based prediction of sandstone thickness is challenging. Inspired by the autoencoder, we develop an open-source deep-learning workflow that combines unsupervised and supervised learning with jointed latent eigenvalues of the convolutional autoencoder (CAE) and traditional seismic attributes for seismic facies classification and sandstone thickness prediction constrained by the facies distribution. First, we extract lower-dimensional latent eigenvalues as a category of novel seismic attributes from the seismic data using a CAE. To accurately and effectively extract lower-dimensional latent eigenvalues, we develop a hybrid loss function based on the mean-squared error loss and the smooth L1 loss in this CAE. Then, we use principal component (PC) analysis to extract the first four PCs of these seismic lower-dimensional latent eigenvalues. Using unsupervised K-means, we cluster the first four PCs to form seismic facies. Finally, we take the first four PCs with the traditional seismic attributes as input and the sandstone thickness as labels for the random forest to predict the sandstone thickness distribution. The results of seismic facies and sandstone thickness distribution confirm the potential and advantages of our workflow, which can speed up the identification of seismic facies with smoother boundaries, improve the prediction accuracy by 16% over than that of traditional seismic attributes, and provide more depositional insight for a turbidite reservoir of the Shahejie Formation in the Niuzhuang Delta, China.

A pluggable single-image super-resolution algorithm based on second-order gradient loss

Convolutional neural networks for single-image super-resolution have been widely used with great success. However, most of these methods use L1 loss to guide network optimization, resulting in blurry restored images with sharp edges smoothed. This is because L1 loss limits the optimization goal of the network to the statistical average of all solutions within the solution space of that task. To go beyond the L1 loss, this paper designs an image super-resolution algorithm based on second-order gradient loss. We impose additional constraints by considering the high-order gradient level of the image so that the network can focus on the recovery of fine details such as texture during the learning process. This helps to alleviate the problem of restored image texture over-smoothing to some extent. During network training, we extract the second-order gradient map of the generated image and the target image of the network by minimizing the distance between them, this guides the network to pay attention to the high-frequency detail information in the image and generate a high-resolution image with clearer edge and texture. Besides, the proposed loss function has good embeddability and can be easily integrated with existing image super-resolution networks. Experimental results show that the second-order gradient loss can significantly improve both Learned Perceptual Image Patch Similarity (LPIPS) and Frechet Inception Distance score (FID) performance over other image super-resolution deep learning models.

Appearance and Pose-Conditioned Human photograph generation using Deformable GANs

In this study, we address the challenge of generating images of individuals based on pose and appearance data. Specifically, we take an image, xa, of an individual and a target pose, P (xb), extracted from another image, xb. We then generate a new image of the same individual in the target pose, P (xb), while preserving the visual details from xa. To manage pixel-to-pixel misalignments caused by pose differences between P (xa) and P (xb), we incorporate deformable skip connections in our Generative Adversarial Network’s generator. Additionally, we propose a nearest-neighbour loss as an alternative to the standard L1 and L2 losses to match the texture of the generated image with the target image. Our approach demonstrates competitive quantitative and qualitative results using standard datasets and protocols recently proposed for this task. We also carry out a comprehensive evaluation using off-the-shelf person-identification (Re-id) systems trained with person-generation-based augmented data, a key application for this task. Our experiments reveal that our Deformable GANs can significantly boost Re-id accuracy, surpassing data-augmentation techniques specifically trained using Re-identification losses. index Terms—Conditional GAN, Image Generation, Deformable Objects, Human Pose.

Satellite image fusion using cyclic spatio-spectral GAN model

ABSTRACT Preserving spectral and spatial information in the satellite image fusion problem is one of the essential challenges. This paper presents a GAN-based method for fusing panchromatic and multispectral satellite images. The proposed method utilises the idea of Cycle-GAN with two generators, for spectral and spatial information preservation, based on residual in residual dense block super-resolution architecture. First, generator-1 translates panchromatic and multispectral satellite images to a high-resolution fused image and then preserves details by utilising generator-2. The goal is to reach the spatial and spectral details of panchromatic and multispectral satellite images, respectively. Two discriminators are employed, one for spectral and the other for spatial transformations, and weighted L1 loss is used for both as cycle losses. By leveraging the unique capabilities of the two generators, the proposed method achieves high-quality image fusion results with improved spectral and spatial resolutions that are evaluated over two different datasets. The experimental results demonstrate the effectiveness of the proposed method, with enhancements ranging from approximately 2% to 30% and 0.5% to 35% in WorldView2 and GeoEye-1 images, respectively, in five metrics including PSNR. Moreover, we show significant enhancements of approximately 7% to 50% and 22% to 29% in the same datasets in metrics like SAM and ERGAS.

Investigating the Impact of L2 Attainment on L1 Attrition: The Case of Saudi Children Studying in International Schools in Saudi Arabia

Despite the significant body of research in linguistics, there has been limited exploration of the attrition or retention of the first language in a homeland after a person ceases to use or learn it. However, this phenomenon has become apparent in the Saudi community, particularly in the younger generation who are losing their mother tongue (L1) despite living in their homeland. To address this issue, the present study focuses on L1 attrition and examines lexical disfluency in the oral production of Arabic among 36 Saudi children living in Saudi Arabia. They are L1 Arabic speakers who learned English as their second language L2 while studying in international schools from grade one to six. The study compares the level of their Arabic proficiency to that of other Saudi children studying in national schools. The latter group speaks Arabic as their L1 and English as a foreign language. The aim of the study is to identify the extent to which exposure to English as L2 affects the oral production of Arabic among L1 Arabic speakers. This study is significant because the loss of L1 can result in the erosion of cultural identity and the weakening of social ties within a community (Genesee, 2008). The study provides insights into the factors that contribute to L1 attrition. This can inform efforts to preserve Arabic language and culture in Saudi Arabia, particularly among younger generations who are at risk of losing their L1. The results showed high level of attrition in the attritor group.