Deep Clustering Research Articles

Deep Fusion Clustering Network With Reliable Structure Preservation.

Deep clustering, which can elegantly exploit data representation to seek a partition of the samples, has attracted intensive attention. Recently, combining auto-encoder (AE) with graph neural networks (GNNs) has accomplished excellent performance by introducing structural information implied among data in clustering tasks. However, we observe that there are some limitations of most existing works: 1) in practical graph datasets, there exist some noisy or inaccurate connections among nodes, which would confuse network learning and cause biased representations, thus leading to unsatisfied clustering performance; 2) lacking dynamic information fusion module to carefully combine and refine the node attributes and the graph structural information to learn more consistent representations; and 3) failing to exploit the two separated views' information for generating a more robust target distribution. To solve these problems, we propose a novel method termed deep fusion clustering network with reliable structure preservation (DFCN-RSP). Specifically, the random walk mechanism is introduced to boost the reliability of the original graph structure by measuring localized structure similarities among nodes. It can simultaneously filter out noisy connections and supplement reliable connections in the original graph. Moreover, we provide a transformer-based graph auto-encoder (TGAE) that can use a self-attention mechanism with the localized structure similarity information to fine-tune the fused topology structure among nodes layer by layer. Furthermore, we provide a dynamic cross-modality fusion strategy to combine the representations learned from both TGAE and AE. Also, we design a triplet self-supervision strategy and a target distribution generation measure to explore the cross-modality information. The experimental results on five public benchmark datasets reflect that DFCN-RSP is more competitive than the state-of-the-art deep clustering algorithms. The corresponding code is available at https://github.com/gongleii/DFCN-RSP.

Segmentation of Thoracic Organs through Distributed Extraction of Visual Feature Patterns Utilizing Resio-Inception U-Net and Deep Cluster Recognition Techniques.

Segmentation of medical images plays a key role in the correct identification and management of different diseases. In this study, we present a new segmentation method that meets the difficulties posed by sophisticated organ shapes in computed tomography (CT) images, particularly targeting lung, breast, and gastric cancers. Our suggested methods, Resio-Inception U-Net and Deep Cluster Recognition (RIUDCR), use a Residual Inception Architecture, which combines the power of residual connections and inception blocks to achieve cutting-edge segmentation performance while reducing the risk of overfitting. We present mathematical equations and functions that describe the design, including the encoding and decoding steps within the UC-Net system. Furthermore, we provide strong testing results that show the effectiveness of our method. Through thorough testing on varied datasets, our method regularly beats current techniques, achieving amazing precision and stability in organ task segmentation. These results show the promise of our residual inception architecture in better medical picture analysis. In summary, our research not only shows a state-of-the-art segment methodology but also reinforces its usefulness through thorough testing. The inclusion of residual inception architecture in medical picture segmentation offers good possibilities for improving the identification and management of disease planning.

Robust deep image clustering using convolutional autoencoder with separable discrete Krawtchouk and Hahn orthogonal moments

By cooperatively learning features and assigning clusters, deep clustering is superior to conventional clustering algorithms. Numerous deep clustering algorithms have been developed for a variety of application levels; however, the majority are still incapable of learning robust noise-resistant latent features, which limits the clustering performance. To address this open research challenge, we introduce, for the first time, a new approach called: Robust Deep Embedded Image Clustering algorithm with Separable Krawtchouk and Hahn Moments (RDEICSKHM). Our approach leverages the advantages of Krawtchouk and Hahn moments, such as local feature extraction, discrete orthogonality, and noise tolerance, to obtain a meaningful and robust image representation. Moreover, we employ LayerNormalization to further improve the latent space quality and facilitate the clustering process. We evaluate our approach on four image datasets: MNIST, MNIST-test, USPS, and Fashion-MNIST. We compare our method with several deep clustering methods based on two metrics: clustering accuracy (ACC) and normalized mutual information (NMI). The experimental results show that our method achieves superior or competitive performance on all datasets, demonstrating its effectiveness and robustness for deep image clustering.

Heterogeneous trading behaviors of individual investors: A deep clustering approach

While individual investors may have more diverse preferences and trading behavior than institutional investors due to their lack of professional education, many studies tend to lump individual investors together or classify them by socio-demographic characteristics. We conducted an empirical study using account-level trading data for over 300,000 investors in the Korean stock market from 2016 to 2020 to analyze the heterogeneity of individual investors. Our findings reveal notable disparities in profit distributions among the clusters formed based on investors' trading behavior. Therefore, this study emphasizes the importance of exploring the heterogeneity of individual investors to understand their behavior better.

Deep Adaptive Fuzzy Clustering for Evolutionary Unsupervised Representation Learning.

Cluster assignment of large and complex datasets is a crucial but challenging task in pattern recognition and computer vision. In this study, we explore the possibility of employing fuzzy clustering in a deep neural network framework. Thus, we present a novel evolutionary unsupervised learning representation model with iterative optimization. It implements the deep adaptive fuzzy clustering (DAFC) strategy that learns a convolutional neural network classifier from given only unlabeled data samples. DAFC consists of a deep feature quality-verifying model and a fuzzy clustering model, where deep feature representation learning loss function and embedded fuzzy clustering with the weighted adaptive entropy is implemented. We joint fuzzy clustering to the deep reconstruction model, in which fuzzy membership is utilized to represent a clear structure of deep cluster assignments and jointly optimize for the deep representation learning and clustering. Also, the joint model evaluates current clustering performance by inspecting whether the resampled data from estimated bottleneck space have consistent clustering properties to improve the deep clustering model progressively. Experiments on various datasets show that the proposed method obtains a substantially better performance for both reconstruction and clustering quality compared to the other state-of-the-art deep clustering methods, as demonstrated with the in-depth analysis in the extensive experiments.

Multitask-Guided Deep Clustering With Boundary Adaptation.

Multitask learning uses external knowledge to improve internal clustering and single-task learning. Existing multitask learning algorithms mostly use shallow-level correlation to aid judgment, and the boundary factors on high-dimensional datasets often lead algorithms to poor performance. The initial parameters of these algorithms cause the border samples to fall into a local optimal solution. In this study, a multitask-guided deep clustering (DC) with boundary adaptation (MTDC-BA) based on a convolutional neural network autoencoder (CNN-AE) is proposed. In the first stage, dubbed multitask pretraining (M-train), we construct an autoencoder (AE) named CNN-AE using the DenseNet-like structure, which performs deep feature extraction and stores captured multitask knowledge into model parameters. In the second phase, the parameters of the M-train are shared for CNN-AE, and clustering results are obtained by deep features, which is termed as single-task fitting (S-fit). To eliminate the boundary effect, we use data augmentation and improved self-paced learning to construct the boundary adaptation. We integrate boundary adaptors into the M-train and S-fit stages appropriately. The interpretability of MTDC-BA is accomplished by data transformation. The model relies on the principle that features become important as the reconfiguration loss decreases. Experiments on a series of typical datasets confirm the performance of the proposed MTDC-BA. Compared with other traditional clustering methods, including single-task DC algorithms and the latest multitask clustering algorithms, our MTDC-BA achieves better clustering performance with higher computational efficiency. Deep features clustering results demonstrate the stability of MTDC-BA by visualization and convergence verification. Through the visualization experiment, we explain and analyze the whole model data input and the middle characteristic layer. Further understanding of the principle of MTDC-BA. Through additional experiments, we know that the proposed MTDC-BA is efficient in the use of multitask knowledge. Finally, we carry out sensitivity experiments on the hyper-parameters to verify their optimal performance.

Deep Spectral Clustering With Constrained Laplacian Rank.

Spectral clustering (SC) is a well-performed and prevalent technique for data processing and analysis, which has attracted significant attention in the field of clustering. While the scalability and generalization ability of this method make it prohibitive for the large-scale dataset and the out-of-sample-extension problem. In this work, we propose a new efficient deep clustering architecture based on SC, named deep SC (DSC) with constrained Laplacian rank (DSCCLR). DSCCLR develops a self-adaptive affinity matrix with a clustering-friendly structure by constraining the Laplacian rank, which greatly mines the intrinsic relationships. Meanwhile, by introducing a simple fully connected network with an orthogonality constraint on the last layer, DSCCLR learns discriminative representations in a short training time. The proposed method has the following salient properties: 1) it overcomes limited generalization ability and scalability of the existing DSC methods; 2) it explores the intrinsic relationship between samples in the affinity matrix, which maintains the latent manifold of data as much as possible; and 3) it alleviates the complexity of eigendecomposition via a simple but effective fully connected network. The extensive empirical results demonstrate the superiorities of DSCCLR over other 17 clustering methods.

A Catalog of 13CO Clumps from the MWISP in l = 10°–20°

In this study, we present a catalog of molecular clumps extracted from 13CO (J = 1 − 0) emission data of the Milky Way Imaging Scroll Painting (MWISP) project. The data covers the inner Milky Way within the longitude range 10° ≤ l ≤ 20° and the latitude strip of ∣b∣ ≤ 5.°25. The workflow for the extraction of clumps, namely Facet-SS-3D-Clump, consists of two parts: the identification of clump candidates and their verification. First, Facet-SS-3D-Clump employs FacetClumps to identify clump candidates. Subsequently, high-confidence clumps are obtained by cross-matching with the clumps detected by other algorithms, such as dendrogram. Second, these high-confidence clumps are used as prior knowledge to train a semi-supervised deep clustering approach, SS-3D-Clump, which is applied to verify clump candidates detected by FacetClumps, providing confidence levels for the molecular clumps. Finally, the catalog comprising 18,757 molecular clumps was obtained using Facet-SS-3D-Clump, and the catalog is 90% complete above 37 K km s−1. We observe a significant deviation of the mean Galactic latitude for clumps within ∣b∣ ≤ 2° from the midplane, with b = − 0.°110. We found that 82.3% of the dust clumps correspond to 13CO clumps by matching with Herschel infrared dust clumps. In the future, Facet-SS-3D-Clump will be applied to detect 13CO clumps in the entire MWISP data.

ScTPC: a novel semisupervised deep clustering model for scRNA-seq data.

Continuous advancements in single-cell RNA sequencing (scRNA-seq) technology have enabled researchers to further explore the study of cell heterogeneity, trajectory inference, identification of rare cell types, and neurology. Accurate scRNA-seq data clustering is crucial in single-cell sequencing data analysis. However, the high dimensionality, sparsity, and presence of "false" zero values in the data can pose challenges to clustering. Furthermore, current unsupervised clustering algorithms have not effectively leveraged prior biological knowledge, making cell clustering even more challenging. This study investigates a semisupervised clustering model called scTPC, which integrates the triplet constraint, pairwise constraint, and cross-entropy constraint based on deep learning. Specifically, the model begins by pretraining a denoising autoencoder based on a zero-inflated negative binomial distribution. Deep clustering is then performed in the learned latent feature space using triplet constraints and pairwise constraints generated from partial labeled cells. Finally, to address imbalanced cell-type datasets, a weighted cross-entropy loss is introduced to optimize the model. A series of experimental results on 10 real scRNA-seq datasets and five simulated datasets demonstrate that scTPC achieves accurate clustering with a well-designed framework. scTPC is a Python-based algorithm, and the code is available from https://github.com/LF-Yang/Code or https://zenodo.org/records/10951780.

Resting-state functional connectivity in children cooled for neonatal encephalopathy.

Therapeutic hypothermia improves outcomes following neonatal hypoxic-ischaemic encephalopathy, reducing cases of death and severe disability such as cerebral palsy compared with normothermia management. However, when cooled children reach early school-age, they have cognitive and motor impairments which are associated with underlying alterations to brain structure and white matter connectivity. It is unknown whether these differences in structural connectivity are associated with differences in functional connectivity between cooled children and healthy controls. Resting-state functional MRI has been used to characterize static and dynamic functional connectivity in children, both with typical development and those with neurodevelopmental disorders. Previous studies of resting-state brain networks in children with hypoxic-ischaemic encephalopathy have focussed on the neonatal period. In this study, we used resting-state fMRI to investigate static and dynamic functional connectivity in children aged 6-8 years who were cooled for neonatal hypoxic-ischaemic without cerebral palsy [n = 22, median age (interquartile range) 7.08 (6.85-7.52) years] and healthy controls matched for age, sex and socioeconomic status [n = 20, median age (interquartile range) 6.75 (6.48-7.25) years]. Using group independent component analysis, we identified 31 intrinsic functional connectivity networks consistent with those previously reported in children and adults. We found no case-control differences in the spatial maps of these intrinsic connectivity networks. We constructed subject-specific static functional connectivity networks by measuring pairwise Pearson correlations between component time courses and found no case-control differences in functional connectivity after false discovery rate correction. To study the time-varying organization of resting-state networks, we used sliding window correlations and deep clustering to investigate dynamic functional connectivity characteristics. We found k = 4 repetitively occurring functional connectivity states, which exhibited no case-control differences in dwell time, fractional occupancy or state functional connectivity matrices. In this small cohort, the spatiotemporal characteristics of resting-state brain networks in cooled children without severe disability were too subtle to be differentiated from healthy controls at early school-age, despite underlying differences in brain structure and white matter connectivity, possibly reflecting a level of recovery of healthy resting-state brain function. To our knowledge, this is the first study to investigate resting-state functional connectivity in children with hypoxic-ischaemic encephalopathy beyond the neonatal period and the first to investigate dynamic functional connectivity in any children with hypoxic-ischaemic encephalopathy.

An overview on deep clustering

In recent years, with the great success of deep learning and especially deep unsupervised learning, many deep architectural clustering methods, collectively known as deep clustering, have emerged. Deep clustering shows the potential to outperform traditional methods, especially in handling complex high-dimensional data, taking full advantage of deep learning. To achieve a comprehensive overview of the field of deep clustering, this review systematically explores deep clustering methods and their various applications. First, the basic network architecture of deep clustering is described in detail, including the common network frameworks, and loss functions. Subsequently, deep clustering is divided into several categories based on the network architecture, and benchmark datasets and evaluation metrics in the field are introduced. Next, the real-world applications of deep clustering are explored in depth, providing successful cases in the fields of bioinformatics, medicine, anomaly detection, and image processing, highlighting the broad applicability of deep clustering in solving real-world challenges. Finally, the paper summarizes its contributions and explores potential directions for future research in deep clustering.

Unsupervised structural damage identification based on covariance matrix and deep clustering

SummaryStructural damage identification is a major task in structural health monitoring. Machine learning and deep learning algorithms have been widely applied in the research of structural damage identification. Supervised algorithms require expert labeling, making it difficult to implement in engineering applications. Unsupervised structural damage identification algorithms are generally divided into two parts: damage‐sensitive factor extraction and damage determination. Existing algorithms all perform these two steps separately. This paper proposes a damage identification method combining covariance matrix and improved deep embedding clustering network (IDEC). IDEC can perform damage‐sensitive factor extraction and damage determination operations at the same time. The covariance matrix that introduces delay information contains rich damage features, and the combination of the two has been proven to effectively mine the damage‐sensitive feature space. After network hyperparameter optimization via Bayesian optimization, the proposed method is applied to the damage identification and quantification using real bridge acceleration response data under vehicle load. The results show that this method can identify structural damage with an accuracy of up to 97% with better performance than existing technologies, and it also has great performance in identifying small damages. The proposed method is expected to increase the damage identification accuracy if applied in engineering practice.

Day-ahead Numerical Weather Prediction solar irradiance correction using a clustering method based on weather conditions

Accurate solar irradiance forecasts can make solar power forecasts more reliable, which can help the power grid dispatch reasonably. In practice, Numerical Weather Prediction (NWP) is widely applied in solar irradiance forecasts. In this paper, the statistical NWP Global Horizontal Irradiance (GHI) error analysis shows that the characteristics of NWP GHI error vary obviously under different weather conditions. However, existing correction methods are not designed contrapuntally for different weather conditions, resulting in poor correction performance. To solve this problem, a hybrid method is proposed to get day-ahead correction results for NWP GHI. Specifically, the hybrid method consists of three key parts, including Deep Clustering (DC), Variational Mode Decomposition (VMD), and an Encoder–Decoder based Correction model (EDC). DC is used to categorize the historical samples into three clusters, the input of which is the multi-dimensional information series containing observed data and NWP data. After a feature selection by VMD, the correction models are trained on each cluster respectively. The performance of the proposed model is evaluated with a public dataset and an actual field dataset, and the results demonstrate that the accuracy has been effectively improved by the proposed method compared with other models. In addition, we find that adopting VMD is effective in improving the correction accuracy, and the root mean square error is reduced by 5.70% and 9.32% compared with models without it.

Robust deep fuzzy [formula omitted]-means clustering for image data

Image clustering is a difficult task with important application value in computer vision. The key to this task is the quality of images features. Most of current clustering methods encounter the challenge. That is, the process of feature learning and clustering operates independently. To address this problem, several researchers have been dedicated to performing feature learning and deep clustering together. However, the obtained features lack discriminability to address high-dimensional data successfully. To deal with this issue, we propose a novel model named as robust deep fuzzy K-means clustering (RD-FKC), which efficiently projects image samples into a representative embedding space and precisely learns membership degrees into a combined framework. Specifically, RD-FKC introduces Laplacian regularization technique to preserve locality properties of data. Moreover, by using an adaptive loss function, the model becomes more robust to diverse types of outliers. Furthermore, to avoid the latent space being distorted and make the extracted features retain the original information as much as possible, the model introduces reconstruction error and adds regularization to network parameters. Finally, an effective algorithm is derived to solve the optimization model. Numerous experiments have been conducted, illustrating the advantages and superiority of RD-FKC over existing clustering approaches.

Integrated Somatic Mutation Network Diffusion Model for Stratification of Breast Cancer into Different Metabolic Mutation Subtypes

Background: Mutations in metabolism-related genes in somatic cells potentially lead to disruption of metabolic pathways, which results in patients exhibiting different molecular and pathological features. background: Mutations in metabolism-related genes in somatic cells potentially lead to disruption of metabolic pathways, which results in patients exhibiting different molecular and pathological features. Objective: In this study, we focused on somatic mutation data to investigate the significance of metabolic mutation typing in guiding the prognosis and treatment of breast cancer patients. objective: In this study, we focused on somatic mutation data to investigate the significance of metabolic mutation typing in guiding the prognosis and treatment of breast cancer patients. Methods: The somatic mutation profile of breast cancer patients was analyzed and smoothed by utilizing a network diffusion model within the protein-protein interaction network to construct a comprehensive somatic mutation network diffusion profile. Subsequently, a deep clustering approach was employed to explore metabolic mutation typing in breast cancer based on integrated metabolic pathway information and the somatic mutation network diffusion profile. In addition, we employed deep neural networks and machine learning prediction models to assess the feasibility of predicting drug responses through somatic mutation network diffusion profiles. Results: Significant differences in prognosis and metabolic heterogeneity were observed among the different metabolic mutation subtypes, characterized by distinct alterations in metabolic pathways and genetic mutations, and these mutational features offered potential targets for subtype-specific therapies. Furthermore, there was a strong consistency between the results of the drug response prediction model constructed on the somatic mutation network diffusion profile and the actual observed drug responses. Conclusion: Metabolic mutation typing of cancer assists in guiding patient prognosis and treatment.

An End-to-end Deep Clustering Method with Consistency and Complementarity Attention Mechanism for Multisensor Fault Diagnosis

Deep clustering methods have found successful applications in single-sensor data fault diagnosis. However, most of these methods employ separate optimization strategies that overlook the interaction between feature learning and clustering. Moreover, conventional deep learning methods for fault diagnosis often disregard the consistent and complementary information inherent in the multisensor data, resulting in unsatisfactory multisensor fault diagnosis performance. In this study, we introduce a novel End-to-end Deep Clustering Method with Consistency and Complementarity Attention Mechanism, termed EDCM-CCAM, tailored for multisensor fault diagnosis. Firstly, multiple deep autoencoder networks are utilized to concurrently extract the deep representation features from various sensor inputs. Secondly, we introduce a Consistency and Complementarity Attention Mechanism (CCAM) to facilitate multisensor feature fusion, accompanied by the design of two distinct loss functions to fully exploit the consistent and complementary information within multisensor data. Finally, fault pattern recognition in multisensor data is accomplished through Kullback-Leibler (KL) divergence-based clustering, while a joint optimization strategy is employed to simultaneously optimize all components of the EDCM-CCAM. The efficacy of the proposed method is validated on a gearbox dataset, demonstrating superior performance in multisensor fault diagnosis compared to alternative methods.

Recognition and optimisation method of impact deformation patterns based on point cloud and deep clustering: Applied to thin-walled tubes

The recognition and clustering of deformation modes are key to constructing impact deformation constraints for thin-walled structures. This paper transforms the clustering and recognition problem of structural impact deformation modes into a problem of clustering and recognition of point cloud sequences based on pseudo-labels. The effectiveness of the method is assessed, and the experimental results show that the accuracy of the proposed method can reach up to 92.17 % when using a pre-training deep neural network feature extractor, which is not only close to the 98.50 % accuracy of supervised learning classification models but also has a 16.84 % improvement in accuracy compared to the deep clustering method based on K-Means. Under different clustering conditions, the proposed method can effectively classify and recognise samples with similar deformation modes and has the ability to summarise and induce new deformation modes when the number of clusters exceeds the number of manual labels. Furthermore, this paper presents a multi-objective optimisation method for structural crashworthiness under impact deformation constraints based on the NSGA-II algorithm. This method constructs impact deformation constraints using surrogate models and deformation clustering and recognition models. The experimental results show that the proposed method can effectively constrain the generation of the population. It is found that there are a large number of Pareto solutions that do not belong to the expected impact deformation mode under the condition of no deformation mode constraint. In contrast, almost all the obtained Pareto solutions conform to the expected impact deformation mode under the condition of deformation mode constraint. In summary, under the condition of impact deformation constraint, the obtained Pareto solutions can satisfy the crashworthiness requirements while conforming to the expected impact deformation mode.

DeepCluE: Enhanced Deep Clustering via Multi-Layer Ensembles in Neural Networks

DeepCluE: Enhanced Deep Clustering via Multi-Layer Ensembles in Neural Networks

Parallelly Adaptive Graph Convolutional Clustering Model.

Benefiting from exploiting the data topological structure, graph convolutional network (GCN) has made considerable improvements in processing clustering tasks. The performance of GCN significantly relies on the quality of the pretrained graph, while the graph structures are often corrupted by noise or outliers. To overcome this problem, we replace the pre-trained and fixed graph in GCN by the adaptive graph learned from the data. In this article, we propose a novel end-to-end parallelly adaptive graph convolutional clustering (AGCC) model with two pathway networks. In the first pathway, an adaptive graph convolutional (AGC) module alternatively updates the graph structure and the data representation layer by layer. The updated graph can better reflect the data relationship than the fixed graph. In the second pathway, the auto-encoder (AE) module aims to extract the latent data features. To effectively connect the AGC and AE modules, we creatively propose an attention-mechanism-based fusion (AMF) module to weight and fuse the data representations of the two modules, and transfer them to the AGC module. This simultaneously avoids the over-smoothing problem of GCN. Experimental results on six public datasets show that the effectiveness of the proposed AGCC compared with multiple state-of-the-art deep clustering methods. The code is available at https://github.com/HeXiax/AGCC.

Exploration on relation between vehicle oscillation type and platoon oscillation evolution based on multi-scenario field experiment

In order to explore the categorization of vehicle oscillation and the relationship between individual vehicle oscillation and platoon oscillation in traffic flow, this paper conducts field experiments with both human driving vehicles(HDVs) and autonomous vehicles(AVs). The objective is to utilize experimental data to categorize and quantify traffic oscillation, delve into the impact of different degrees of vehicle oscillation on platoon oscillation, identify patterns between vehicle oscillation types and platoon oscillation evolution, and study the influence of AVs on platoon oscillation. The paper commences with car-following experiments involving HDVs and AVs, sets the platoon into three scenarios of HDV’s platoon, AV’s platoon, and mixed platoon experimental datasets. Subsequently, nine oscillation features reflecting the oscillation characteristics of vehicle traveling, such as volatility and trend, etc., are extracted from the experimental datasets. Then the CLS deep clustering model is constructed, trained, and outputs high-dimensional features. The model's effectiveness is assessed using both clustering error and deep learning error simultaneously. Finally, the paper analyzes the classification results of the experimental dataset and explores the relationship between vehicle oscillation type and platoon oscillation. The research finding indicates that the oscillation classification effectiveness of the CLS model is significantly superior to other algorithms. Using the CLS model, vehicle oscillations are categorized into four most suitable types: I, II, III, and IV, representing slight oscillation, general oscillation, relatively serious oscillation, and severe oscillation respectively. In HDVs’ platoon, as the vehicle oscillation type increases, the following vehicle does not fully adhere to the front vehicle, the amplitude trend of the vehicle changes, leading to oscillation in the following vehicle platoon. In the AVs’ platoon, AVs suppress the backward propagation of oscillations in the platoon, and the oscillation types are all type I, indicating mild oscillation. In the mixed platoon, it is observed that AVs, through swift perception of the state of preceding vehicle, promptly respond to adjust their own driving condition to alleviate platoon oscillation. The higher the proportion of AVs in the platoon, the stronger the inhibitory effect on platoon oscillation.