Analyzing the fast search and find of density peaks clustering (DPC) algorithm, we that cluster centers cannot be determined automatically selected may fall into a local optimum random selection parameter cut-off distance dc value. To overcome these problems, novel algorithm based on DPC & PSO (PDPC) is proposed. Particle swarm optimization (PSO) introduced because its simple concept strong global ability, which can optimal solution in relatively few iterations. First, to solve effect...

The gas-phase metallicity is a crucial parameter for understanding the evolution of galaxies. Considering that number multiband galaxy images can typically reach tens millions, using these as input data to predict has become feasible method. However, accuracy estimates from relatively limited. To solve this problem, we propose measurement residual network (GPM-ResNet), deep learning method designed photometric DESI. parameters are labeled with values, which were obtained through...

The cluster number can directly affect the clustering effect and its application in real-world scenarios. Its determination is one of key issues analysis. According to singular value decomposition (SVD), characteristic directions larger values likely represent primary data patterns, trends, or structures corresponding main information. In analysis, information structure are related itself. may correspond clusters, their different clusters. Based on this, a singular-value-based detection...

Prototypes help to explain the predictions of deep classification models for time series. However, most learn prototypes by randomly initializing an uncertain number low-discriminative prototypes, which may lead unstable and unreliable results. To address these issues, we propose a new class Discriminative Prototype Learning Network (DPL-Net), learns appropriate class-discriminative thus improving performance. Specifically, proposed Initialization Mechanism (PIM) introduces proximity metric...

Abstract Carbon stars are chemically peculiar with high carbon abundance, showing strong molecule bands and rare emission lines in their spectra. This paper explores the stellar activity of by identifying line features An outlier detection method based on morphological feature extraction interval representation is used to identify 88 targets presenting from 3546 star Of these, 55 present Balmer series emissions, 35 show forbidden ([O i ] λ 6301 Å, [O iii ], [N ii [S only 2 spectra that not...

The missing satellite problem remains a central issue of the Lambda cold dark matter (ΛCDM) model. On small scale, number observed dwarf galaxies is still fewer than predicted by existing theories. Therefore, finding fainter in deeper images crucial for refining theoretical framework. In this study, we propose an end-to-end automatic identification scheme size and faint based on Dark Energy Spectroscopic Instrument (DESI) Legacy Imaging Surveys photometric images, provide batch galaxy...

As one of the most important techniques in data mining, clustering has always been highly concerned. Most algorithms have encountered challenges, such as difficulty cluster centers selection, artificial determination number clusters K, low accuracy clustering, and uneven efficiency different sets. Considering chosen, a new algorithm fast selecting initial is proposed this paper. Generally, are those points with higher density, smaller radius threshold far away from each other, method uses...

The extensive growth of data quantity has posed many challenges to analysis and retrieval. Noise redundancy are typical representatives the above-mentioned challenges, which may reduce reliability retrieval results increase storage computing overhead. To solve above problems, a two-stage pre-processing framework for noise identification reduction, called ARIS, is proposed in this article. first stage identifies removes noises by following steps: First, influence space (IS) introduced...

Rough set models have been widely used as a method for feature selection in fault diagnosis. A neighbourhood rough model can deal with both nominal and numerical features, but selecting the size its application may be challenge. In this article, authors illustrate that using common all features overestimate or underestimate feature's dependency degree. The is then modified by setting different sizes features. applied to diagnosis of slurry pump impellers. Results show chosen subsets...

In this study, we discover that the data skewness problem imposes adverse impacts on MapReduce-based parallel kNN-join operations running clusters. We propose a partitioning approach-called kNN-DP-to alleviate load imbalance incurred by skewness. The overarching goal of kNN-DP is to equally divide objects into large number partitions, which are processed mappers and reducers in parallel. At heart module, dynamically judiciously partitions optimize performance suppressing Hadoop Data...

Traditional abnormal trajectory detection algorithms mainly involve the measurement of a single feature; however, influence other features on is ignored, resulting in inability to fully discover database. To overcome this limitation, we propose an method - called TADSS find hidden by using comprehensive measurement. Firstly, employ three kernel functions measure time, velocity and position feature values data, where extract semantic position, time trajectory, object motion from each data....

LAMOST (Large Sky Area Multi-Object Fiber Spectroscopic Telescope) has completed the observation of nearly 20 million celestial objects, including a class spectra labeled `Unknown'. Besides low signal-to-noise ratio, these often show some anomalous features that do not work well with current templates. In this paper, total 638,000 `Unknown' from DR5 are selected, and an unsupervised-based analytical framework named SA-Frame (Spectra Analysis-Frame) is provided to explore their origins...

Stellar spectral template libraries play an important role in the automated analysis of stellar spectra. Synthetic cover a very large parameter space but suffer from poor matching with observed In this study, we propose synthetic-to-observed translation (SOST) method based on generative adversarial networks. The SOST is able to calibrate synthetic spectra by converting them corresponding We applied Kurucz and data Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST). After...

Effective extraction of data association rules can provide a reliable basis for classification stellar spectra. The concept spectrum weighted itemsets and are introduced, the weight single property in is determined by information entropy. On that basis, method presented to mine based on frequent pattern tree. Important properties spectral line highlighted using this method. At same time, waveform whole taken into account. experimental results show mined with consistent main features types.