Cluster Parameters Research Articles

Deep Adaptive Graph Clustering via von Mises-Fisher Distributions

Graph clustering has been a hot research topic and is widely used in many fields, such as community detection in social networks. Lots of works combining auto-encoder and graph neural networks have been applied to clustering tasks by utilizing node attributes and graph structure. These works usually assumed the inherent parameters (i.e., size and variance) of different clusters in the latent embedding space are homogeneous, and hence the assigned probability is monotonous over the Euclidean distance between node embeddings and centroids. Unfortunately, this assumption usually does not hold since the size and concentration of different clusters can be quite different, which limits the clustering accuracy. In addition, the node embeddings in deep graph clustering methods are usually L2 normalized so that it lies on the surface of a unit hyper-sphere. To solve this problem, we proposed D eep A daptive G raph C lustering via von Mises-Fisher distributions, namely DAGC. DAGC assumes the node embeddings H can be drawn from a von Mises-Fisher distribution and each cluster k is associated with cluster inherent parameters ρ k which includes cluster center μ and cluster cohesion degree κ. Then we adopt an EM-like approach (i.e., 𝒫( H | ρ ) and 𝒫( ρ | H ), respectively) to learn the embedding and cluster inherent parameters alternately. Specifically, with the node embeddings, we proposed to update the cluster centers in an attraction-repulsion manner to make the cluster centers more separable. And given the cluster inherent parameters, a likelihood-based loss is proposed to make node embeddings more concentrated around cluster centers. Thus, DAGC can simultaneously improve the intra-cluster compactness and inter-cluster heterogeneity. Finally, extensive experiments conducted on four benchmark datasets have demonstrated that the proposed DAGC consistently outperforms the state-of-the-art methods, especially on imbalanced datasets.

Prediction of Pathological Risk Stratification using Computed Tomography Features in Gastrointestinal Stromal Tumours: A Retrospective observational Study

Introduction: Gastrointestinal Stromal Tumours (GISTs) are the most common mesenchymal tumours of gastrointestinal tract. A high postsurgical recurrence and metastatic rate have created a need for a presurgical risk profile identification system. Aim: To assess the association between morphological Computed Tomography (CT) parameters with the pathological risk profile and analyse which CT features can predict the risk grading of GISTs. Materials and Methods: This was a retrospective cohort study based on imaging and histopathological data of 26 patients with pathologically proven GISTs presenting to the Department of Radiodiagnosis of a tertiary hospital in the northern Indian Himalayan foothills over a period of five years from July 2018 to June 2023. CT imaging features including size, growth pattern, margins, enhancement, calcifications, necrosis, intralesional haemorrhage, enlarged feeding vessels, direct organ invasion, and associations such as ascites and lymphadenopathy were studied. All lesions were classified as per Miettinen risk classification into no risk, very low-risk, low, moderate, and high-risk lesions. Analysis was done by the Chi-square test. Predictive analysis was carried out by computing the odds ratio and performing regression analysis on significantly associated imaging features. Results: Out of 26 patients, the study group comprised 16 males (61.54%) and 10 females (38.46%). The most common decade of presentation was the 6th decade with the mean age of presentation being 55.81±4.23 years. Twelve patients were grouped under intermediate to high-risk grading. Lesion size >5 cm (p-value=0.0171, OR=19.12), ill-defined margins (p-value=0.0048, OR=18.33), intralesional necrosis (p-value=0.0053, OR=19.8), and enlarged feeding vessels (p-value=0.012, OR=21.27) were identified as imaging features with significant association and predictive ability for high-risk lesions. The strongest predictive ability for a high-risk profile was shown by ill-defined margins (R2 =0.381) and intralesional necrosis (R2 =0.3287). Conclusion: A preoperative Contrast Enhanced Computed Tomography (CECT) assessment provides a comprehensive imaging profile for GISTs as well as a fair accuracy of risk profile prediction via various singular and clustered morphological parameters.

A Data‐Driven Method for Supporting Self‐Adapt Large‐Scale Group Decision‐Making: A Case Study on Resilient Design of Firm’s Product

Large‐scale group decision‐making (LSGDM) has emerged as a prominent research area in various domains, such as high technology and complex engineering problems. The advent of machine learning techniques has revolutionized LSGDM by introducing new data‐driven approaches. First, recurrent neural networks (RNNs) have been proposed as a data‐driven method to effectively learn and predict experts’ preferences. Second, a self‐adaptive method has been devised to optimize clustering parameters, considering their influence. The consensus‐reaching process facilitates the reverse optimization of these parameters. Third, a novel approach called analysis target cascading (ATC) has been suggested to address the limitations of traditional weighing methods used in previous LSGDM studies. ATC comprehensively investigates the potential game among multiple subgroups, thereby resolving the consensus‐reaching problem (CRP). Lastly, an improved artificial bee colony algorithm has been proposed to tackle the optimization problem presented in this study. This enhanced algorithm incorporates the levying mechanism and searching method from the gravity search algorithm. To validate the efficacy of the proposed methods, a case study involving a large‐scale interdisciplinary team has been conducted.

Sleep-Disordered Breathing: Statistical Characteristics Of Joint Recurrent Indicators In ЕЕG Activity

The purpose of this study was to identify promising candidates for the role of biomarkers associated with different degrees of the apnea-hypapnea index in patients using polysomnographic recordings. Material — The study used polysomnography data recorded in 30 patients with nocturnal respiratory dysfunction in the form of obstructive sleep apnea syndrome. Methods — Analysis of polysomnographic recordings was carried out using a joint recurrent indicator, for which further statistical characteristics were assessed: average value, geometric mean, cubic mean, median, dispersion, standard deviation, the coefficient of variation, asymmetry indicator, kurtosis indicator. Results — For all polysomnographic recordings, joint recurrence diagrams were calculated to identify time points corresponding to specific sleep events in patients with high and low apnea-hypnea index. Based on the statistical characteristics of such events, possible candidates for the role of biomarkers to diagnose apnea syndrome are introduced. Conclusion — The article presents clustering parameters and the efficiency of dividing into clusters of statistical characteristics for two groups of patients - with high and low apnea-hypnea index. Characteristics have been identified that are promising candidates for the role of biomarkers associated with the apnea-hypnea index value.

DataColor: unveiling biological data relationships through distinctive color mapping.

In the era of rapid advancements in high-throughput omics technologies, the visualization of diverse data types with varying orders of magnitude presents a pressing challenge. To bridge this gap, we introduce DataColor, an all-encompassing software solution meticulously crafted to address this challenge. Our aim is to empower users with the ability to handle a wide array of data types through an assortment of tools, while simultaneously streamlining parameter selection for rapid insights and detailed enhancements. DataColor stands as a robust toolkit, encompassing 23 distinct tools coupled with over 600 parameters. The defining characteristic of this toolkit is its adept utilization of the color spectrum, allowing for the representation of data spanning diverse types and magnitudes. Through the integration of advanced algorithms encompassing data clustering, normalization, squarified layouts, and customizable parameters, DataColor unveils an abundance of insights that lay hidden within the intricate relationships embedded in the data. Whether you find yourself navigating the analysis of expansive datasets or embarking on the quest to visualize intricate patterns, DataColor stands as the comprehensive and potent solution. We extend the availability of DataColor to all users at no cost, accessible through the following link: https://github.com/frankgenome/DataColor.

Comparative studies of Markov chain-based driving cycles for light-duty vehicles

ABSTRACT With the massive ongoing efforts to mitigate climate change and reduce emission factors in the transportation sector, driving cycles (DCs) are becoming an essential tool for the testing and certification of distinct types of vehicles. Nevertheless, the open literature lacks careful comparative studies on the impact of different approaches in the development of Markov chain (MC)-based DCs, as well as DCs, developed specifically for the developing and highly congested urban areas in the Middle East and North Africa. Using a large dataset of 43 light-duty vehicles, driven over different areas in Greater Cairo, Egypt, this study aims to develop, compare, and benchmark 24 candidate MC-based DCs, against two clustering-based DCs, as well as four cycles used widely in the US and Europe. These 24 DCs differ in terms of the clustering algorithm (K-medoids and K-means), clustering parameters (different combinations of vehicle’s speed, acceleration, specific power, and percentage idling time), and definitions of microtrips (start-stop and fixed distance). The results show that the MC method outperforms random chaining of microtrips, with average relative root mean square errors (RRMSEs) of 15.8% and 23.6%, respectively. Clustering 350 m fixed distance-based microtrips using the vehicle’s speed, acceleration, and percentage idling time shows the least RRMSE of 8.207%. Defining microtrips based on fixed distance is also better than starts/stops for most vehicle types. The reference cycles (WLTP, NEDC, UDDS, and FTP-17) showed poor representativeness of the real-world data, with an average RRMSE of 76.8%. The same inferior performance of reference cycles, compared to the newly proposed ones, was also highlighted in the estimation of fuel consumption and emission factors. Hence, the proposed cycles and the reported comparative studies can be valuable tools for the assessment of emissions and fuel consumption in such developing metropolitan areas.

Parameter Estimation for Open Clusters using an Artificial Neural Network with a QuadTree-based Feature Extractor

With the unprecedented increase in the number of known star clusters, quick and modern tools are needed for their analysis. In this work, we develop an artificial neural network (ANN) trained on synthetic clusters to estimate the age, metallicity, extinction, and distance of Gaia open clusters. We implement a novel technique to extract features from the color–magnitude diagram of clusters by means of the QuadTree tool, and we adopt a multiband approach. We obtain reliable parameters for ∼5400 clusters. We demonstrate the effectiveness of our methodology in accurately determining crucial parameters of Gaia open clusters by performing a comprehensive scientific validation. In particular, with our analysis we have been able to reproduce the Galactic metallicity gradient as it is observed by high-resolution spectroscopic surveys. This demonstrates that our method reliably extracts information on metallicity from color–magnitude diagrams (CMDs) of stellar clusters. For the sample of clusters studied, we find an intriguing systematic older age compared to previous analyses present in the literature. This work introduces a novel approach to feature extraction using a QuadTree algorithm, effectively tracing sequences in CMDs despite photometric errors and outliers. The adoption of ANNs, rather than convolutional neural networks, maintains the full positional information and improves performance, while also demonstrating the potential for deriving cluster parameters from simultaneous analysis of multiple photometric bands, beneficial for upcoming telescopes like the Vera Rubin Observatory. The implementation of ANN tools with robust isochrone fit techniques could provide further improvements in the quest for open cluster parameters.

TWO TOOLS FOR ESTIMATING STAR CLUSTER PARAMETERS

A star cluster is a composition of stars held together by the overall gravitational field of the whole cluster. The stars of a given cluster are born in a giant molecular cloud and, as such, can be regarded as objects with almost equal ages. Furthermore, assuming that the initial material in the cloud is perfectly mixed, we may also say that the metallicity of all cluster members is the same. Four parameters (distance, extinction or reddening, age, and metallicity) are standardly used to describe star clusters. Their knowledge is essential for studying galaxies’ local and global properties (especially our own Galaxy). However, deriving these parameters may take time and effort. Thanks to the Gaia mission’s parallax measurements, we can determine the distances of stars (and clusters) within our Galaxy with unprecedented precision. Therefore, we can remove the distance from the list of free parameters. We developed two different tools for estimating cluster parameters: 1) Metalcode, an automatic tool focused on deriving metallicities of open clusters, and 2) Stellar Isochrone Fitting Tool (StIFT), a tool for fitting a grid of isochrones for any photometric system available. We present both tools in more detail.

Climate analysis of wave systems for multimodal sea states in the Mediterranean Sea

We present a new methodology for the detection of directional wave systems at a given location. Our approach leverages image segmentation algorithms to group integral spectral parameters in homogeneous clusters, represented by as many centroids associated to specific wave directions and capable to summarize the modes of the local wave climate. The method was tested in the Mediterranean Sea taking advantage of a long-term hindcast dataset, and highlighted well-known climatic patterns consistent to previous research in the area. By identifying the prevailing wave directions, the method allows to synthesize information useful for a plethora of ocean and coastal engineering applications. Among these, we quantify the frequency of occurrence of crossing seas in the Mediterranean Sea.

The impact of halo concentration on the Sunyaev Zel’dovich effect signal from massive galaxy clusters

ABSTRACT The Sunyaev Zel’dovich (SZ) effect is sensitive to the pressure of ionized gas inside galaxy clusters, which is in turn controlled largely by the gravitational potential of the cluster. Changing the concentration parameter describing the cluster mass distribution impacts the gravitational potential and thus the cluster SZ signal, with implications for cosmological and other analyses of SZ-selected clusters. We investigate the concentration–SZ relation in theory and simulations. We find that the impact of concentration on the inner SZ profile (R ≲ 0.75R200c) can be captured with standard polytropic gas models. However, we find that such models do a poor job of reproducing the outer SZ profiles (R ≳ 0.75R200c) and the relation between the integrated SZ signal, Y, and concentration. This disagreement results from a sharp truncation of the gas pressure profile near the splashback radius, likely caused by virial shocks. We develop a simple description of the truncation that leads to a good match with the simulated SZ profiles out to several R200c for clusters of varying mass and concentration, and that also accurately predicts the concentration–Y relationship. Finally, we determine how inference of the linear bias parameter and splashback radius for SZ-selected clusters can be biased by ignoring the concentration dependence of the SZ signal, finding that bias to the former is essentially negligible, while bias to the latter can be as much as 2 per cent.

Is the M81 Fast Radio Burst Host Globular Cluster Special?

We use multiband archival Hubble Space Telescope (HST) observations to measure the photometric and structural parameters of the M81 globular cluster (GC) that hosts the fast radio burst FRB 20200120E. Our best-fitting King model has an effective radius r h = 3.06 pc with a moderate King model concentration of c = 53, and an inferred core radius of 0.81 pc. We revisit the exact astrometric location of the FRB within the cluster, and find that FRB 20200120E is located 1.92 pc from the center, but within the projected half-light radius. We estimate the relative encounter rate of the FRB host, along with the corresponding rates of 210 other GCs in M81, and compare these values with the encounter rates of Galactic GCs. The FRB resides in a GC with an encounter rate that is moderately higher than the median stellar encounter rate in our two comparison samples. While the estimated encounter rate of the FRB host cluster (e.g., ∼50% of a cluster like 47 Tuc) is sufficient to allow the possibility that the FRB formed dynamically, our results do not place strong constraints on this scenario due to the limitations of the available HST data and the possible systematic uncertainties and selection effects in the comparison data.

Electronic Structure of Small Isolated and Supported Manganese Oxide Clusters.

In the present work, possible molecular models of the isolated manganese oxides and supported Mn3Ox/Al2O3 structures were built based on small clusters of passivated MnOx. The support was represented as a simplified model of the alumina tetramer cluster based on small fragments of AlOxHy. Combinations of MnOxHy and AlOxHy clusters were made to form both the isolated and supported manganese oxides clusters. The electronic structure of these systems was characterized by ab initio methods (DFT and CASPT2). It was observed that the vertical excitation energy of the isolated and supported Mn3OxHy clusters is significantly lower than that of the alumina cluster model, while both the isolated and supported Mn3OxHy wave function characters are qualitatively similar with respect to the ground state and electronic transition processes, suggesting that the alumina cluster behaves as an inert support, since there is little contribution of this component in the description of the low-lying electronic states. The present study also reports for the first time the spectroscopic parameters of several clusters containing the manganese transition metal atom.

STRP-DBSCAN: A Parallel DBSCAN Algorithm Based on Spatial-Temporal Random Partitioning for Clustering Trajectory Data

Trajectory clustering algorithms analyze the movement trajectory of the target objects to mine the potential movement trend, regularity, and behavioral patterns of the object. Therefore, the trajectory clustering algorithm has a wide range of applications in the fields of traffic flow analysis, logistics and transportation management, and crime analysis. Existing algorithms do not make good use of the temporal attributes of trajectory data, resulting in a long clustering time and low clustering accuracy of spatial-temporal trajectory data. Meanwhile, the density-based clustering algorithms represented by DBSCAN are very sensitive to the clustering parameters. The radius value Eps and the minimal points number MinPts within Eps radius, defined by the user, have a significant impact on the clustering results, and tuning these parameters is difficult. In this paper, we present STRP-DBSCAN, a parallel DBSCAN algorithm based on spatial-temporal random partitioning for clustering trajectory data. It adopts spatial-temporal random partitioning to distribute balanced computation among different computing nodes and reduce the communication overhead of the parallel clustering algorithm, thus improving the execution efficiency of the DBSCAN algorithm. We also present the PER-SAC algorithm, which uses deep reinforcement learning to combine the prioritized experience replay (PER) and the soft actor-critic (SAC) algorithm for autotuning the optimal parameters of DBSCAN. The experimental results show that STRP-DBSCAN effectively reduces the clustering time of spatial-temporal trajectory data by up to 96.2% and 31.2% compared to parallel DBSCAN and the state-of-the-art RP-DBSCAN. The PER-SAC algorithm also outperforms the state-of-the-art DBSCAN parameter tuning algorithms and improves the clustering accuracy by up to 8.8%. At the same time, the proposed algorithm obtains a higher stability of clustering accuracy.

Clustering Properties of Intermediate and High-mass Young Stellar Objects* * Table is also available online at the CDS in the VizieR archive service.

We have selected 337 intermediate- and high-mass young stellar objects (YSOs; 1.5–20 M ⊙) well-characterized with spectroscopy. By means of the clustering algorithm HDBSCAN, we study their clustering and association properties in the Gaia DR3 catalog as a function of stellar mass. We find that the lower-mass YSOs (1.5–4 M ⊙) have clustering rates of 55%–60% in Gaia astrometric space, a percentage similar to that found in the T Tauri regime. However, intermediate-mass YSOs in the range 4–10 M ⊙ show a decreasing clustering rate with stellar mass, down to 27%. We find tentative evidence suggesting that massive YSOs (>10 M ⊙) often (yet not always) appear clustered. We put forward the idea that most massive YSOs form via a mechanism that demands many low-mass stars around them. However, intermediate-mass YSOs form in a classical core-collapse T Tauri way, yet they do not appear often in the clusters around massive YSOs. We also find that intermediate- and high-mass YSOs become less clustered with decreasing disk emission and accretion rate. This points toward an evolution with time. For those sources that appear clustered, no major correlation is found between their stellar properties and the cluster sizes, number of cluster members, cluster densities, or distance to cluster centers. In doing this analysis, we report the identification of 55 new clusters. We tabulated all of the derived cluster parameters for the considered intermediate- and high-mass YSOs.

Multidimensional investigation of the effect of pre-treatment solutions on drying characteristics and raisin quality

This study aims to evaluate the relationships between pre-treatment solution, variety, drying characteristics, and raisin quality in raisin production, using multivariate analysis methods. This study was conducted on raisins obtained by dipping Bineteti and Zeyti local seed grape varieties in 13 different pre-treatment solutions and drying in the sun on a concrete drying platform. In the study, data of 15 numerical variables related to drying characteristics and raisin quality were reduced to four principal components (PC1, PC2, PC3 and PC4) by principal component analysis (PCA), and their score values were numerically obtained. Then, two grape varieties, 13 pre-treatment solutions, and the four principal components were analyzed by non-linear principal component analysis (NLPCA). In addition, cluster analysis was performed to determine the prominent pre-treatment solutions in terms of drying characteristics and raisin quality. It was determined that the pre-treatment solutions were effective on L*, a*, b*, chroma (C*), hue (h°), a/b values, antioxidant activity, total phenolic content, and drying time constituting PC1. It was remarkable that the colour parameters in prominent clusters in cluster analysis also form PC1 in PCA analysis. The best pre-treatment solutions were found to be the "5% K2CO3 + 1% olive oil" solution for the Bineteti variety and the "5% K2CO3 + 2% hazelnut oil" solution for the Zeyti variety. It was determined that the pre-treatment solutions recommended for the varieties caused to be obtained brighter, more vibrant, and yellow-coloured raisins and shortened the drying time, and had positive effects on the total phenolic content and antioxidant activity.

Variable stars in the field of the Galactic bulge globular cluster NGC 6522

We present a comprehensive analysis of the variable stars projected on the field of the Galactic bulge globular cluster NGC 6522, offering valuable insights into their characteristics. Using proper motions from Gaia-DR3, we aim to distinguish between field stars and true cluster members. For an accurate color-magnitude diagram of the member stars, we produced a differential reddening map. We detect and discuss the peculiarities of variable stars of the types RR Lyrae, type II Cepheids, Long Period Variables (LPV) and eclipsing binaries, whose light curves are available through the OGLE III and IV databases. Notably, we explore the variable V24, which shows a prominent phase modulation resulting from period changes in a time scale of 1100 days. The variable stars among the cluster members serve as indicators of the cluster metallicity and distance; these determinations are based on their light curves. With the Fourier light curve decomposition of three RRc stars, we have derived the following cluster parameters: the metallicity in the spectroscopic scale text{[Fe/H]}_{mathrm{UVES}}=-1.16 pm 0.09; and the mean distance D=8.77 pm 0.16 kpc.

An XFEM-based CZM Numerical Strategy for Modeling Hydraulic Fracture Propagation under Different Fracturing Schemes

Segmented and cluster fracturing can improve the efficiency of volume stimulation in shale reservoirs and reduce construction costs. It is important to clarify the propagation characteristics of hydraulic fractures and stress interference under different fracturing techniques to optimize the process of clustered and staged fracturing. For this purpose, we have developed a 2D XFEM-based CZM hydraulic fracturing model. The capability of this model was validated by analytical solutions and then used to study the propagation paths of hydraulic fractures and the characteristics of stress interference under simultaneous fracturing, sequential fracturing, zipper fracturing and multi-cluster fracturing. The results show that in simultaneous fracturing, the middle cluster is compressed by the external position clusters, and the opening width of hydraulic fractures is reduced. In sequential fracturing, the fracture that first initiates create an additional stress field that inhibits the propagation distance of subsequent fractures and the propagation path of hydraulic fractures is also affected in the stress shadow region. Zipper fracturing can effectively alleviate stress interference between multiple fractures, and internal fractures can also propagate a certain distance. In multi-cluster fracturing, the fluid rate into the internal fractures may be limited, and fracture propagation is also limited by stress interference. Therefore, it is necessary to optimize the parameters of clusters to ensure that all clusters can initiate fractures normally. The research results are important for the parameter optimization of clustered and staged fracturing, especially for well factory fracturing mode.

Wafer Image Preprocessing Based on Density Based Spatial Clustering of Application with Noise

With the development of microelectronic manufacturing technology, semiconductor manufacturing presents the trend of maximization of scale and miniaturization of process size. Even if the wafer production now has a high automation of the production process, high-precision production equipment and advanced production technology, wafer abnormal situation is still inevitable. Abnormal conditions in semiconductor manufacturing process will reduce the yield of wafer products and increase production costs. The later analysis becomes the necessary means to improve the wafer yield. With the rapid development of modern computing power, the application of machine learning-based automatic detection method in semiconductor production is irresistible. In addition to its spatial pattern, there are many noises that can affect the classification of defects, so it is necessary to preprocess the wafer diagram. The traditional density based spatial clustering of application with noise (DBSCAN) algorithm needs to determine two clustering parameters artificially, and the choice of parameters can affect the clustering effect easily, the parameter list was obtained by K-mean nearest neighbor algorithm and mathematical expectation method, and the integrated parameters of intra-cluster density and inter-cluster density after DBSCAN clustering were selected as evaluation indexes to select the optimal parameters. Finally, a self-adaptive (SA)-DBSCAN map is obtained by retaining the largest cluster, adding feature points and feature clusters, thus improving the yield of wafer.

Do the early galaxies observed by JWST disagree with Planck's CMB polarization measurements?

The recent observations from the James Webb Space Telescope have led to a surprising discovery of a significant density of massive galaxies with masses of M ≥ 1010.5 M ⊙ at redshifts of approximately z ∼ 10. This corresponds to a stellar mass density of roughly ρ * ∼ 106 M ⊙ Mpc-3. Despite making conservative assumptions regarding galaxy formation, this finding may not be compatible with the standard ΛCDM cosmology that is favored by observations of CMB Anisotropies from the Planck satellite. In this paper, we confirm the substantial discrepancy with Planck's results within the ΛCDM framework. Assuming a value of ϵ = 0.2 for the efficiency of converting baryons into stars, we indeed find that the ΛCDM model is excluded at more than 99.7 % confidence level (C.L.). An even more significant exclusion is found for ϵ ∼ 0.1, while a better agreement, but still in tension at more than 95 %, is obtained for ϵ = 0.32.This tension, as already discussed in the literature, could arise either from systematics in the JWST measurements or from new physics. Here, as a last-ditch effort, we point out that disregarding the large angular scale polarization obtained by Planck, which allows for significantly larger values of the matter clustering parameter σ 8, could lead to better agreement between Planck and JWST within the ΛCDM framework. Assuming ΛCDM and no systematics in the current JWST results, this implies either an unknown systematic error in current large angular scale CMB polarization measurements or an unidentified physical mechanism that could lower the expected amount of CMB polarization produced during the epoch of reionization. Interestingly, the model compatible with Planck temperature-only data and JWST observation also favors a higher Hubble constant H 0 = 69.0±1.1 km/s/Mpc at 68% C.L., in better agreement with observations based on SN-Ia luminosity distances.

Dual Clustering Co-Teaching With Consistent Sample Mining for Unsupervised Person Re-Identification

In unsupervised person Re-ID, peer-teaching strategy leveraging two networks to facilitate training has been proven to be an effective method to deal with the pseudo label noise. However, training two networks with a set of noisy pseudo labels reduces the complementarity of the two networks and results in label noise accumulation. To handle this issue, this paper proposes a novel Dual Clustering Co-teaching (DCCT) approach. DCCT mainly exploits the features extracted by two networks to generate two sets of pseudo labels separately by clustering with different parameters. Each network is trained with the pseudo labels generated by its peer network, which can increase the complementarity of the two networks to reduce the impact of noises. Furthermore, we propose dual clustering with dynamic parameters (DCDP) to make the network adaptive and robust to dynamically changing clustering parameters. Moreover, Consistent Sample Mining (CSM) is proposed to find the samples with unchanged pseudo labels during training for potential noisy sample removal. Extensive experiments demonstrate the effectiveness of the proposed method, which outperforms the state-of-the-art unsupervised person Re-ID methods by a considerable margin and surpasses most methods utilizing camera information.