Semi-supervised Learning Method Research Articles

A Semi-Supervised Transformer-Based Deep Learning Framework for Automated Tooth Segmentation and Identification on Panoramic Radiographs.

Automated tooth segmentation and identification on dental radiographs are crucial steps in establishing digital dental workflows. While deep learning networks have been developed for these tasks, their performance has been inferior in partially edentulous individuals. This study proposes a novel semi-supervised Transformer-based framework (SemiTNet), specifically designed to improve tooth segmentation and identification performance on panoramic radiographs, particularly in partially edentulous cases, and establish an open-source dataset to serve as a unified benchmark. A total of 16,317 panoramic radiographs (1589 labeled and 14,728 unlabeled images) were collected from various datasets to create a large-scale dataset (TSI15k). The labeled images were divided into training and test sets at a 7:1 ratio, while the unlabeled images were used for semi-supervised learning. The SemiTNet was developed using a semi-supervised learning method with a label-guided teacher-student knowledge distillation strategy, incorporating a Transformer-based architecture. The performance of SemiTNet was evaluated on the test set using the intersection over union (IoU), Dice coefficient, precision, recall, and F1 score, and compared with five state-of-the-art networks. Paired t-tests were performed to compare the evaluation metrics between SemiTNet and the other networks. SemiTNet outperformed other networks, achieving the highest accuracy for tooth segmentation and identification, while requiring minimal model size. SemiTNet's performance was near-perfect for fully dentate individuals (all metrics over 99.69%) and excellent for partially edentulous individuals (all metrics over 93%). In edentulous cases, SemiTNet obtained statistically significantly higher tooth identification performance than all other networks. The proposed SemiTNet outperformed previous high-complexity, state-of-the-art networks, particularly in partially edentulous cases. The established open-source TSI15k dataset could serve as a unified benchmark for future studies.

Semi-supervised segmentation of medical images focused on the pixels with unreliable predictions

Pseudo-labeling is a well-studied approach in semi-supervised learning. However, unreliable or potentially incorrect pseudo-labels can accumulate training errors during iterative self-training steps, leading to unstable performance. Addressing this challenge typically involves either discarding unreliable pseudo-labels, resulting in the loss of important data, or attempting to refine them, risking the possibility of worsening the pseudo-labels in some cases/pixels. In this paper, we propose a novel method based on pseudo-labeling for semi-supervised segmentation of medical images. Unlike existing approaches, our method neither discards any data nor worsens reliable pseudo-labels. Our approach generates uncertainty masks for the predictions, utilizing reliable pixels without any modification as ground truths and modifying the unreliable ones rather than discarding them. Furthermore, we introduce a novel loss function that incorporates both mentioned parts by multiplying each term by its corresponding uncertainty mask, encompassing reliable and unreliable pixels. The reliable pixels are addressed using a masked cross-entropy loss function, while the modification of the unreliable pixels is performed through a deep-learning-based adaptation of active contours. The entire process is solved within a single loss function without the need to solve traditional active contour equations. We evaluated our approach on three publicly available datasets, including MRI and CT images from cardiac structures and lung tissue. Our proposed method outperforms the state-of-the-art semi-supervised learning methods on all three datasets. Implementation of our work is available at https://github.com/behnam-rahmati/Semi-supervised-medical.

Borehole lithology modelling with scarce labels by deep transductive learning

Geophysical logging is a geo-scientific instrument that detects information such as electric, acoustic, and radioactive properties of a well. Its data plays a vital role in interpreting subsurface geology. However, since logging data is an indirect reflection of rocks, it requires the construction of a logging interpretation model in combination with core samples. Obtaining and analysing all core samples in a well is not practical due to their enormous cost, leading to the problem of scarce core sample labels. This problem can be addressed using semi-supervised learning. Existing studies on lithology identification using logging data mostly utilize graph-based semi-supervised learning, which requires known features to establish a graph Laplacian matrix. Therefore, these methods often use logging values at certain depths to construct feature vectors and cannot learn the shape information of logging curves. In this paper, we propose a semi-supervised learning method with feature learning capability based on semi-supervised generative adversarial network (SSGAN) to learn the shape information of logging curves while utilizing unlabelled logging curves. Additionally, considering the problem of insufficient use of labels when dividing a validation set in extremely scarce-label situations, we propose a strategy of weighted averaging of three sub-models, which effectively improves model performance. We verify the effectiveness of our proposed method on five wells and demonstrate the mechanism of semi-supervised learning using adversarial learning through extensive visualization methods.

Semi-supervised noise-resilient anomaly detection with feature autoencoder

Most methods only use normal samples to learn anomaly detection (AD) models in an unsupervised manner. However, these samples may be noisy in real-world applications, causing the models to be unable to accurately identify anomaly objects. In addition, there are a small number of anomaly samples in real industrial production that should be fully utilized to help model discrimination. Existing methods of introducing anomaly samples still have bottlenecks in model identification capabilities. In this paper, by introducing both normal and a few abnormal samples, we propose a novel semi-supervised learning method for anomaly detection, named RobustPatch, which can improve the model discriminability through a self-cross scoring mechanism and the learning of feature AutoEncoder. Our approach contains two core designs: Firstly, we propose a self-cross scoring module, calculating the weights of normal and anomaly features extracted from corresponding images using a self-scoring and cross-scoring manner, respectively. Secondly, our approach proposes a fully connected feature AutoEncoder to rate the extracted features, which is trained with the supervision of the scored weights. Extensive experiments on the MVTecAD and BTAD datasets validate the superior anomaly boundaries discriminability of our approach and superior performance in noise-polluted scenarios.

Mussel culture monitoring with semi-supervised machine learning on multibeam echosounder data using label spreading

High diversity seabed habitats, such as shellfish aggregations, play a significant role in marine ecosystem sustainability but are susceptible to bottom disturbance induced by anthropogenic activities. Regular monitoring of these habitats with effective mapping methods is therefore essential. Multibeam echosounder (MBES) has been widely used in recent decades for seabed characterization due to its non-destructive manner and extensive spatial coverage compared to traditional methods like bottom sampling. Nevertheless, bottom sampling remains essential to link ground truth with acoustic seabed classification. Using seabed samples and MBES measurements, machine learning techniques are commonly employed to model their relationships and generate classification maps of an extended seabed. However, limited ground truth data, resulting from constraints in regulations, budget, or time, may impede the development of robust machine learning models. To address this challenge, we applied a semi-supervised machine learning method to classify seabed sediments of a blue mussel (Mytilus edulis) cultivation area in the Oosterschelde, the Netherlands. We utilized nine boxcore samples to generate pseudo-labels on MBES data. These pseudo-labels enlarged the training data size, facilitated the training of three comprehensive machine learning algorithms (Gradient Boosting, Random Forest, and Support Vector Machine), and helped to classify the study site into mussel and non-mussel areas. We found the geomorphological and backscatter-related features to be complementary for mussel culture detection. Our classification results were demonstrated effective through expert knowledge of this cultivation area and brought insights for future research on natural mussel habitats.

A Comparative Analysis of U-Net and Vision Transformer Architectures in Semi-Supervised Prostate Zonal Segmentation.

The precise segmentation of different regions of the prostate is crucial in the diagnosis and treatment of prostate-related diseases. However, the scarcity of labeled prostate data poses a challenge for the accurate segmentation of its different regions. We perform the segmentation of different regions of the prostate using U-Net- and Vision Transformer (ViT)-based architectures. We use five semi-supervised learning methods, including entropy minimization, cross pseudo-supervision, mean teacher, uncertainty-aware mean teacher (UAMT), and interpolation consistency training (ICT) to compare the results with the state-of-the-art prostate semi-supervised segmentation network uncertainty-aware temporal self-learning (UATS). The UAMT method improves the prostate segmentation accuracy and provides stable prostate region segmentation results. ICT plays a more stable role in the prostate region segmentation results, which provides strong support for the medical image segmentation task, and demonstrates the robustness of U-Net for medical image segmentation. UATS is still more applicable to the U-Net backbone and has a very significant effect on a positive prediction rate. However, the performance of ViT in combination with semi-supervision still requires further optimization. This comparative analysis applies various semi-supervised learning methods to prostate zonal segmentation. It guides future prostate segmentation developments and offers insights into utilizing limited labeled data in medical imaging.

A Semi-Supervised Active Learning Method for Structured Data Enhancement with Small Samples

In order to solve the problems of the small capacity of structured data and uneven distribution among classes in machine learning tasks, a supervised generation method for structured data called WAGAN and a cyclic sampling method named SACS (Semi-supervised and Active-learning Cyclic Sampling), based on semi-supervised active learning, are proposed. The loss function and neural network structure are optimized, and the quantity and quality of the small sample set are enhanced. To enhance the reliability of generating pseudo-labels, a Semi-supervised Active learning Framework (SAF) is designed. This framework redistributes class labels to samples, which not only enhances the reliability of generated samples but also reduces the influence of noise and uncertainty on the generation of false labels. To mine the diversity information of generated samples, an uncertain sampling strategy based on spatial overlap is designed. This strategy incorporates the idea of spatial overlap and uses global and local sampling methods to calculate the information content of generated samples. Experimental results show that the proposed method performs better than other data enhancement methods on three different datasets. Compared to the original data, the average F1macro value of the classification model is improved by 11.5%, 16.1%, and 19.6% relative to compared methods.

Semi-supervised abdominal multi-organ segmentation by object-redrawing.

Multi-organ segmentation is a critical task in medical imaging, with wide-ranging applications in both clinical practice and research. Accurate delineation of organs from high-resolution 3D medical images, such as CT scans, is essential for radiation therapy planning, enhancing treatment outcomes, and minimizing radiation toxicity risks. Additionally, it plays a pivotal role in quantitative image analysis, supporting various medical research studies. Despite its significance, manual segmentation of multiple organs from 3D images is labor-intensive and prone to low reproducibility due to high interoperator variability. Recent advancements in deep learning have led to several automated segmentation methods, yet many rely heavily on labeled data and human anatomy expertise. In this study, our primary objective is to address the limitations of existing semi-supervised learning (SSL) methods for abdominal multi-organ segmentation. We aim to introduce a novel SSL approach that leverages unlabeled data to enhance the performance of deep neural networks in segmenting abdominal organs. Specifically, we propose a method that incorporates a redrawing network into the segmentation process to correct errors and improve accuracy. Our proposed method comprises three interconnected neural networks: a segmentation network for image segmentation, a teacher network for consistency regularization, and a redrawing network for object redrawing. During training, the segmentation network undergoes two rounds of optimization: basic training and readjustment. We adopt the Mean-Teacher model as our baseline SSL approach, utilizing labeled and unlabeled data. However, recognizing significant errors in abdominal multi-organ segmentation using this method alone, we introduce the redrawing network to generate redrawn images based on CT scans, preserving original anatomical information. Our approach is grounded in the generative process hypothesis, encompassing segmentation, drawing, and assembling stages. Correct segmentation is crucial for generating accurate images. In the basic training phase, the segmentation network is trained using both labeled and unlabeled data, incorporating consistency learning to ensure consistent predictions before and after perturbations. The readjustment phase focuses on reducing segmentation errors by optimizing the segmentation network parameters based on the differences between redrawn and original CT images. We evaluated our method using two publicly available datasets: the beyond the cranial vault (BTCV) segmentation dataset (training: 44, validation: 6) and the abdominal multi-organ segmentation (AMOS) challenge 2022 dataset (training:138, validation:16). Our results were compared with state-of-the-art SSL methods, including MT and dual-task consistency (DTC), using the Dice similarity coefficient (DSC) as an accuracy metric. On both datasets, our proposed SSL method consistently outperformed other methods, including supervised learning, achieving superior segmentation performance for various abdominal organs. These findings demonstrate the effectiveness of our approach, even with a limited number of labeled data. Our novel semi-supervised learning approach for abdominal multi-organ segmentation addresses the challenges associated with this task. By integrating a redrawing network and leveraging unlabeled data, we achieve remarkable improvements in accuracy. Our method demonstrates superior performance compared to existing SSL and supervised learning methods. This approach holds great promise in enhancing the precision and efficiency of multi-organ segmentation in medical imaging applications.

Semi-supervised learning for efficient water leakage segmentation in tunnel infrastructure

Water leakage segmentation based on computer vision allows for efficient monitoring and maintenance, ensuring the safety and integrity of tunnel infrastructure. In this study, a semisupervised learning method with self-training via pseudo-labels is proposed to overcome the dependence of conventional computer vision models on high-quality labeled data. Strong data augmentation is first injected into the unlabeled data to establish a robust self-training baseline and differentiate similar predictions between teacher and student networks. To mitigate the accumulation of erroneous pseudo-labels and their potential impact on performance, we chose to retrain using only reliable, unlabeled images. The unlabeled images are prioritized based on their overall stability. By considering image-level contextual information, the selection process offers a more suitable approach for segmentation compared to conventional pixelwise methods. The experimental results highlight the method’s capability to enhance model performance with a limited amount of labeled data, surpassing the effectiveness of other semisupervised methods.

Research on Lithology Identification based on Machine Learning

Machine learning has great potential in lithology identification. Through supervised learning, unsupervised learning, semi-supervised learning, deep learning and other methods, features can be automatically extracted from complex seismic data and logging data to achieve efficient and accurate lithology classification. These methods not only improve the accuracy and efficiency of lithology identification, but also reduce the workload of geologists, allowing them to focus on higher-level analysis and decision making. Despite significant progress, machine learning still faces many challenges in lithology identification. First, data quality and quantity limitations remain a major problem, especially in certain areas where obtaining high-quality seismic and logging data is difficult. Second, the training and interpretability of complex models also need to be addressed, especially because the "black box" nature of deep learning models makes it difficult for geologists to understand their internal mechanisms and predictions. In addition, the generalization ability and overfitting of models, as well as the need for real-time data processing in practical applications, are also urgent challenges to be solved. To address these challenges, this paper proposes several solutions and future directions. Data enhancement and synthesis techniques can extend existing data sets and improve the robustness and accuracy of models. The development of interpretative models and visualization tools helps geologists understand and trust the decision-making process of models. Multi-source data fusion technology can effectively use multi-source heterogeneous data such as seismic data, well logging data and geological map to improve model performance. Online learning and transfer learning technologies can update models in real time, improve the adaptability and generalization ability of models, and develop more accurate and interpretable models.

Semi-Supervised Left-Atrial Segmentation Based on Squeeze–Excitation and Triple Consistency Training

Convolutional neural networks (CNNs) have achieved remarkable success in fully supervised medical image segmentation tasks. However, the acquisition of large quantities of homogeneous labeled data is challenging, making semi-supervised training methods that rely on a small amount of labeled data and pseudo-labels increasingly popular in recent years. Most existing semi-supervised learning methods, however, underestimate the importance of the unlabeled regions during training. This paper posits that these regions may contain crucial information for minimizing the model’s uncertainty prediction. To enhance the segmentation performance of the left-atrium database, this paper proposes a triple consistency segmentation network based on the squeeze-and-excitation mechanism (SETC-Net). Specifically, the paper constructs a symmetric architectural unit called SEConv, which adaptively recalibrates the feature responses in the channel direction by modeling the inter-channel correlations. This allows the network to adaptively weigh each channel according to the task’s needs, thereby emphasizing or suppressing different feature channels. Moreover, SETC-Net is composed of an encoder and three slightly different decoders, which convert the prediction discrepancies among the three decoders into unsupervised loss through a constructed iterative pseudo-labeling scheme, thus encouraging consistent and low-entropy predictions. This allows the model to gradually capture generalized features from these challenging unmarked regions. We evaluated the proposed SETC-Net on the public left-atrium (LA) database. The proposed method achieved an excellent Dice score of 91.14% using only 20% of the labeled data. The experiments demonstrate that the proposed SETC-Net outperforms seven current semi-supervised methods in left-atrium segmentation and is one of the best semi-supervised segmentation methods on the LA database.

GraphixMatch: Improving semi-supervised learning for graph classification with FixMatch

Graph classification is a critical problem that predicts the properties of an entire graph representing complicated relationships. This problem has been extensively studied in a supervised manner in various industries via graph convolutional networks that required large numbers of labeled graphs. However, labeled graphs are typically expensive and time-consuming to obtain in the real-world. Training graph classification models is difficult because of the sparseness of labeled data. Therefore, semi-supervised learning (SSL) algorithms that make full use of unlabeled data and overcome the lack of labels have led to the expansion of graph classification studies. In this study, we propose GraphixMatch to learn the graph properties of both labeled and unlabeled graphs based on the use of FixMatch for graph classification. To effectively capture the characteristics of a graph, we newly define and apply weak and strong augmentation of the graph. We evaluate the proposed method on seven benchmarks from a public dataset. We show that compared with various SSL methods, the simplified proposed method with weak and strong augmentation is superior. Furthermore, we conduct extensive additional experiments to derive the important parameters of GraphixMatch.

A topological approach for semi-supervised learning

Nowadays, Machine Learning and Deep Learning methods have become the state-of-the-art approach to solve data classification tasks. In order to use those methods, it is necessary to acquire and label a considerable amount of data; however, this is not straightforward in some fields, since data annotation is time consuming and might require expert knowledge. This challenge can be tackled by means of semi-supervised learning methods that take advantage of both labelled and unlabelled data. In this work, we present new semi-supervised learning methods based on techniques from Topological Data Analysis (TDA). In particular, we have created two semi-supervised learning methods following two topological approaches. In the former, we have used a homological approach that consists in studying the persistence diagrams associated with the data using the bottleneck and Wasserstein distances. In the latter, we have considered the connectivity of the data. In addition, we have carried out a thorough analysis of the developed methods using 9 tabular datasets with low and high dimensionality. The results show that the developed semi-supervised methods outperform the results obtained with models trained with only manually labelled data, and are an alternative to other classical semi-supervised learning algorithms.

Rapid identification method for on-road high-emission vehicle based on deep semi-supervised anomaly detection

This study aims to identify high-emission vehicles in urban traffic management. Existing high-emission vehicle identification models typically overlook the dispersed anomaly data distribution, and machine learning anomaly detection methods face difficulties in learning decision boundaries, leading to reduced detection performance. To address these issues, this study proposes a semi-supervised learning method based on data fusion and deep anomaly detection. This method integrates the chassis and engine dynamometer test data with on-road remote sensing system data to obtain more comprehensive vehicle emission information. Experimental results demonstrate that the proposed method introduces a penalty mechanism for anomaly samples, encouraging the model to increase the dissimilarity in similarity between normal and abnormal data at the latent data distribution level. In vehicle emission datasets from different regions, this method achieves over 95% AUC, demonstrating strong applicability and accuracy.

Virtual Category Learning: A Semi-Supervised Learning Method for Dense Prediction With Extremely Limited Labels.

Due to the costliness of labelled data in real-world applications, semi-supervised learning, underpinned by pseudo labelling, is an appealing solution. However, handling confusing samples is nontrivial: discarding valuable confusing samples would compromise the model generalisation while using them for training would exacerbate the issue of confirmation bias caused by the resulting inevitable mislabelling. To solve this problem, this paper proposes to use confusing samples proactively without label correction. Specifically, a Virtual Category (VC) is assigned to each confusing sample in such a way that it can safely contribute to the model optimisation even without a concrete label. This provides an upper bound for inter-class information sharing capacity, which eventually leads to a better embedding space. Extensive experiments on two mainstream dense prediction tasks - semantic segmentation and object detection, demonstrate that the proposed VC learning significantly surpasses the state-of-the-art, especially when only very few labels are available. Our intriguing findings highlight the usage of VC learning in dense vision tasks.

Supervised, semisupervised, and unsupervised learning of the Domany-Kinzel model.

The Domany-Kinzel (DK) model encompasses several types of nonequilibrium phase transitions, depending on the selected parameters. We apply supervised, semisupervised, and unsupervised learning methods to studying the phase transitions and critical behaviors of the (1+1)-dimensional DK model. The supervised and the semisupervised learning methods permit the estimations of the critical points, the spatial and temporal correlation exponents, concerning labeled and unlabeled DK configurations, respectively. Furthermore, we also predict the critical points by employing principal component analysis and autoencoder. The PCA and autoencoder can produce results in good agreement with simulated stationary particle number density.

Dirty road extraction from GF-2 images by semi-supervised deep learning method for arid and semiarid regions of southern Mongolia

ABSTRACT The uncontrolled proliferation of natural roads in arid regions has exacerbated regional land degradation and desertification, presenting substantial challenges to their accurate mapping owing to their dynamic and obscure features. Moreover, the high cost of data annotation restricts the availability of comprehensively labelled datasets, which are essential for advanced remote sensing processing and natural road detection. This study dedicated to implement a semi-supervised deep learning method for dirty road extraction in southern Mongolia. A new thematic semantic segmentation dataset of natural roads was established firstly to address scarcity of annotation datasets this region. A semi-supervised UniMatch structure was designed consequently. Operating with high-resolution GaoFen-2 images, this approach minimises the need for extensive manual annotation, achieving an IOU of 73.51% and MIOU of 86.37%. This method significantly reduces labour and time costs associated with manual and fully supervised methods. These observations provide a valuable data source and methodology for addressing natural road expansion in arid regions. They can aid governments in evaluating transportation infrastructure in remote areas, and analysing dirty road traffic impact on environment.

Extraction of Sunspots from Chinese Sunspot Drawings Based on Semisupervised Learning

China has six observing stations, providing over 52,000 handwritten sunspot drawings from 1947–2016. The observing stations are the Purple Mountain Astronomical Observatory (PMO), Yunnan Astronomical Observatory (YNAO), Qingdao Observatory Station (QDOS), Sheshan Observatory Station (SSOS), Beijing Planetarium (BJP), and Nanjing University (NJU). In this paper, we propose a new cotraining semisupervised learning method combining a semantic segmentation method named dynamic mutual training (DMT) boundary-guided semantic segmentation (BGSeg), i.e., DMT_BGSeg, which makes full use of the labeled data from PMO and the unlabeled data from the other five stations to detect and segment sunspot components in all sunspot drawings of the six Chinese stations. The sunspot is detected and segmented. Additionally, each sunspot is split into four types of components: pore, spot, umbra, and hole. The testing results show the mIoU values of PMO, YNAO, BJP, NJU, QDOS and SSOS are 85.29, 72.65, 73.82, 64.28, 62.26, and 60.07, respectively. The results of the comparison also show that DMT_BGSeg is effective in detecting and segmenting sunspots in Chinese sunspot drawings. The numbers and areas of sunspot components are measured separately. All of the detailed data are publicly shared on China-VO, which will advance the comprehensive augmentation of the global historical sunspot database and further the understanding of the long-term solar activity cycle and solar dynamo.

Few-shot learning with long-tailed labels

Few-Shot Learning (FSL) is a challenging classification task in machine learning, and it aims to recognize unseen examples of new classes with only a few labeled reference examples (i.e., the support set). The training phase of FSL typically requires a large amount of labeled examples (i.e., the base set) to effectively learn transferable knowledge, but it is usually difficult to obtain sufficient data annotation in practical applications. Existing semi-supervised FSL approaches can learn generalizable representations from partly labeled data, yet they do not sufficiently consider the real distribution of those labeled data. In this paper, we propose a new problem setting termed Few-Shot Learning with Long-Tailed Labels (FSL-LTL) to further consider a more practical semi-supervised scenario where the labeled examples are long-tailed. To effectively address this new problem, we build a novel two-stage training framework dubbed Reweighted Contrastive Embedding (RCE). In the first stage of RCE, we adopt the popular contrastive learning framework to pre-train a reliable network in a self-supervised manner. In the second stage, we integrate the semi-supervised empirical risk into a Weighted Random Sampling (WRS) strategy to fine-tune the pre-trained backbone with the aid of a consistency regularization. Experimental results demonstrate the feasibility of the proposed FSL-LTL problem setting and the superiority of our new RCE method over existing FSL approaches and semi-supervised learning methods. These results also suggest that the RCE approach is a promising solution for addressing the new FSL-LTL problem.

Identifying tectonic settings of porphyry copper deposits using zircon trace elements – A semi-supervised machine learning method

Porphyry copper deposits (PCDs), as main sources of Cu resources worldwide, have attracted great attention for both scientific research and mining purposes in the past decades. The formation of PCDs is closely related to magmatic-hydrothermal system formed in both (continental/oceanic) arc and non-arc environments. Many low-dimensional geochemistry discriminants were developed for identifying the tectonic environments where ore-related magmas formed. However, long-term geological processes may obscure the geochemistry information recorded in rocks and minerals, which inevitably introduces uncertainties for identifying tectonic environments when using the low-dimensional discriminant diagrams. Here, we proposed a semi-supervised machine learning (ML) approach based on zircon chemistry data (spanning 16 elements) to identify the tectonic environments of PCDs. The results show that the semi-supervised models perform better in identifying the tectonic environments compared to the conventional low-dimensional discriminants. Moreover, as another benefit of big data analysis, ML model can clarify the systematic geochemistry differences of igneous zircons formed in different tectonic environments. The results suggest that (U+Th)/Yb and LREE/HREE are the top two important zircon chemical features for distinguishing PCDs formed in arc or non-arc environments. Specifically, zircons from non-arc show higher (U+Th)/Yb compared to that of continental arc, while zircons from island arc show lowest (U+Th)/Yb and LREE/HREE. Finally, the semi-supervised learning model was applied to the Dexing porphyry copper deposit to approach the longstanding debate regarding the underlying metallogenic setting. The results support the viewpoint that the Dexing porphyry copper deposit was formed in an intracontinental (non-arc) environment related to asthenosphere upwelling.