Stochastic Neighbor Embedding Algorithm Research Articles

Predicting Alloying Element Yield in Converter Steelmaking Using t-SNE-WOA-LSTM

The performance and quality of steel products are significantly impacted by the alloying element control. The efficiency of alloy utilization in the steelmaking process was directly related to element yield. This study analyses the factors that influence the yield of elements in the steelmaking process using correlation analysis. A yield prediction model was developed using a t-distributed stochastic neighbor embedding (t-SNE) algorithm, a whale optimization algorithm (WOA), and a long short-term memory (LSTM) neural network. The t-SNE algorithm was used to reduce the dimensionality of the original data, while the WOA optimization algorithm was employed to optimize the hyperparameters of the LSTM neural network. The t-SNE-WOA-LSTM model accurately predicted the yield of Mn and Si elements with hit rates of 71.67%, 96.67%, and 99.17% and 57.50%, 89.17%, and 97.50%, respectively, falling within the error range of ±1%, ±2%, and ±3% for Mn and ±1%, ±3%, and ±5% for Si. The results demonstrate that the t-SNE-WOA-LSTM model outperforms the backpropagation (BP), LSTM, and WOA-LSTM models in terms of prediction accuracy. The model was applied to actual production in a Chinese plant. The actual performance of the industrial application is within a ±3% error range, with an accuracy of 100%. Furthermore, the elemental yield predicted by the model and then added the ferroalloys resulted in a reduction in the elemental content of the product by 0.017%. The model enables accurate prediction of alloying element yields and was effectively applied in industrial production.

Power quality disturbances classification with imbalanced/insufficient samples based on WGAN-GP-SA and DCNN

The accurate identification of power quality disturbance signals (PQDs) is of great importance to ensure the safe and stable operation of power systems. Considering that there are insufficient samples and imbalance samples of power quality disturbance signals in real systems, the hybrid classification method based on WGAN-GP-SA (Wasserstein generative adversarial networks with gradient penalty and Squeeze-and-Excitation block with atrous spatial pyramid pooling module) and DCNN (deep convolutional neural network) is proposed for recognition of PQDs. Firstly, time-frequency features are extracted by combining improved Kalman and continuous wavelet transforms (CWT); secondly, the original training time-frequency maps are expanded by using WGAN-GP-SA, which can learn the distribution of real data and generate similar samples, and PSNR, SSIM, FID and other indexes are introduced for evaluation of generated data; finally, classification experiments are conducted on two data sets using DCNN, and simulation results show that the accuracy is improved by 2.82 % compared with the original data without augmentation. Furthermore, t-distributed stochastic neighbor embedding algorithm (t-SNE) is introduced for visually analysing of classification performance of various datasets, and the visualization indicates the superiority of the hybrid classification method.

Comparative single-cell multiplex immunophenotyping of therapy-naive patients with rheumatoid arthritis, systemic sclerosis, and systemic lupus erythematosus shed light on disease-specific composition of the peripheral immune system.

Systemic autoimmune diseases (SADs) are a significant burden on the healthcare system. Understanding the complexity of the peripheral immunophenotype in SADs may facilitate the differential diagnosis and identification of potential therapeutic targets. Single-cell mass cytometric immunophenotyping was performed on peripheral blood mononuclear cells (PBMCs) from healthy controls (HCs) and therapy-naive patients with rheumatoid arthritis (RA), progressive systemic sclerosis (SSc), and systemic lupus erythematosus (SLE). Immunophenotyping was performed on 15,387,165 CD45+ live single cells from 52 participants (13cases/group), using an antibody panel to detect 34 markers. Using the t-SNE (t-distributed stochastic neighbor embedding) algorithm, the following 17 main immune cell types were determined: CD4+/CD57- T cells, CD4+/CD57+ T cells, CD8+/CD161- T cells, CD8+/CD161+/CD28+ T cells, CD8dim T cells, CD3+/CD4-/CD8- T cells, TCRγ/δ T cells, CD4+ NKT cells, CD8+ NKT cells, classic NK cells, CD56dim/CD98dim cells, B cells, plasmablasts, monocytes, CD11cdim/CD172dim cells, myeloid dendritic cells (mDCs), and plasmacytoid dendritic cells (pDCs). Seven of the 17 main cell types exhibited statistically significant frequencies in the investigated groups. The expression levels of the 34 markers in the main populations were compared between HCs and SADs. In summary, 59 scatter plots showed significant differences in the expression intensities between at least two groups. Next, each immune cell population was divided into subpopulations (metaclusters) using the FlowSOM (self-organizing map) algorithm. Finally, 121 metaclusters (MCs) of the 10 main immune cell populations were found to have significant differences to classify diseases. The single-cell T-cell heterogeneity represented 64MCs based on the expression of 34 markers, and the frequency of 23 MCs differed significantly between at least twoconditions. The CD3- non-T-cell compartment contained 57 MCs with 17 MCs differentiating at least two investigated groups. In summary, we are the first to demonstrate the complexity of the immunophenotype of34markers over 15 million single cells in HCs vs. therapy-naive patients withRA, SSc, and SLE. Disease specific population frequencies or expression patterns of peripheral immune cells provide a single-cell data resource to the scientific community.

The pistachio quality detection based on deep features plus unsupervised clustering

AbstractThe opening condition of the pistachio is a trait that has been sorted by machine vision. This study presented a qualitative and quantitative method for pistachio opening detection based on unsupervised learning. First, the convolution neural network ResNet18 was utilized to extract the pistachio features based on the different layers: bn4b_branch2b (bn4b), bn5b_branch2b (bn5b), and fc1000. The high‐dimensional feature was mapped to the low‐dimensional by t‐distributed stochastic neighbor embedding algorithm, and then, the clustering algorithms were utilized to cluster the image feature under different spatial dimensions. It was best that the clustering performance came from the bn5b feature layer. The accuracy of this method was 99.31% in three‐dimensional space. And completed the analysis of pistachio opening degree. This method exhibits significant potential for enhancing the quality detection and product classification of pistachios. The unsupervised clustering is utilized to sort not only pistachios but also general nuts.Practical ApplicationsIn this study, we proposed a quality detection method for pistachios based on unsupervised learning, which was capable of performing both qualitative and quantitative analyses of pistachio opening. This method provides an efficient quality detection and product classification strategy, with important practical value. The unsupervised clustering is utilized to sort not only pistachios but also general nuts. Furthermore, it significantly reduces the consumption of manpower and material resources, paving the way for the development of intelligent sorting equipment in the future, and has reference significance for the quality detection of other agricultural products.

Machine learning prediction of mafic–ultramafic rock-related Cr-spinel formation environments and its application to the tectonic settings of magmatic sulfide deposits

This study uses the random forest machine learning algorithm to classify and predict Cr-spinel formation environments in mafic–ultramafic rocks. Cr-spinel is an early-crystallized oxide phase in these rocks and exhibits significant compositional variations that provide valuable insights into their formation environments. Traditionally, empirical diagrams have been used to classify the origins of Cr-spinel based on major and trace element compositions; however, overlap exists among spinel from different formation environments. To overcome this limitation, we develop a random forest model using a compiled dataset of Cr-spinel compositions from various tectonic settings, including small intrusions from continental collisional zones, layered intrusions and small intrusions from intra-continental plates (including large igneous provinces and intra-continental rifts), Alaskan-type complexes from oceanic subduction zones, and ophiolites from oceanic spreading centers. The model is trained, validated, and optimized using cross-validation and hyperparameter tuning. The results show high accuracy in predicting the Cr-spinel formation environments, as demonstrated by high precision, recall, and F1 scores. Feature importance analyses reveal that elements, such as Ti, Cr, and Al, play a more significant role in the prediction. Visualization techniques, i.e., the t-distributed stochastic neighbor embedding (TSNE) algorithm, are applied to reduce the dimensionality of the dataset and create decision boundaries for better interpretation. The model is validated using an independent Cr-spinel composition dataset obtained from an area with known geological settings, to confirm the model’s reliability. The model is applied successfully to reveal the heterogeneous sources of the rifting-related Daxueshan magmatic Ni-Cu sulfide deposit in the Baoshan block in the eastern part of the Tethyan orogenic belt. The machine learning algorithm can accurately distinguish different Cr-spinel types using large and diverse datasets to provide valuable insights into geological formation conditions. This approach is a powerful tool for understanding Earth’s geological processes, improving exploration models of magmatic sulfide deposit and reducing mineral exploration costs.

Noise‐robust gas path fault detection and isolation for a power generation gas turbine based on deep residual compensation extreme learning machine

AbstractOne of the major challenges facing fault diagnosis tools is their exposure to noise. The presence of noise may cause false alarms or the inability to detect a progressive fault in the early stages of its occurrence. Continuing previous efforts to address such a problem, in this paper, a noise‐robust diagnosis system for an industrial gas turbine is presented. The proposed structure employs a set of deep residual compensation extreme learning machines (DRCELMs). In this model, an optimal number of compensating blocks are trained to recover some of the lost useful information in the face of noise. Training and testing data required to develop the fault diagnosis model are generated by a performance model of the studied gas turbine. The t‐distributed stochastic neighbor embedding algorithm is employed for visualizing the gas path faults. Furthermore, the performance of the DRCELM is evaluated by comparing it with six other diagnosis models. The results indicate higher robustness of the DRCELM compared to other fault diagnosis systems. The proposed model presents a classification accuracy of >97% in noisy data and an accuracy of >98% in noise‐free data and combined data, while the average of fault positive rate and fault negative rate in noisy data is less than 2.5%.

A physics-based machine learning technique rapidly reconstructs the wall-shear stress and pressure fields in coronary arteries.

With the global rise of cardiovascular disease including atherosclerosis, there is a high demand for accurate diagnostic tools that can be used during a short consultation. In view of pathology, abnormal blood flow patterns have been demonstrated to be strong predictors of atherosclerotic lesion incidence, location, progression, and rupture. Prediction of patient-specific blood flow patterns can hence enable fast clinical diagnosis. However, the current state of art for the technique is by employing 3D-imaging-based Computational Fluid Dynamics (CFD). The high computational cost renders these methods impractical. In this work, we present a novel method to expedite the reconstruction of 3D pressure and shear stress fields using a combination of a reduced-order CFD modelling technique together with non-linear regression tools from the Machine Learning (ML) paradigm. Specifically, we develop a proof-of-concept automated pipeline that uses randomised perturbations of an atherosclerotic pig coronary artery to produce a large dataset of unique mesh geometries with variable blood flow. A total of 1,407 geometries were generated from seven reference arteries and were used to simulate blood flow using the CFD solver Abaqus. This CFD dataset was then post-processed using the mesh-domain common-base Proper Orthogonal Decomposition (cPOD) method to obtain Eigen functions and principal coefficients, the latter of which is a product of the individual mesh flow solutions with the POD Eigenvectors. Being a data-reduction method, the POD enables the data to be represented using only the ten most significant modes, which captures cumulatively greater than 95% of variance of flow features due to mesh variations. Next, the node coordinate data of the meshes were embedded in a two-dimensional coordinate system using the t-distributed Stochastic Neighbor Embedding (-SNE) algorithm. The reduced dataset for -SNE coordinates and corresponding vector of POD coefficients were then used to train a Random Forest Regressor (RFR) model. The same methodology was applied to both the volumetric pressure solution and the wall shear stress. The predicted pattern of blood pressure, and shear stress in unseen arterial geometries were compared with the ground truth CFD solutions on "unseen" meshes. The new method was able to reliably reproduce the 3D coronary artery haemodynamics in less than 10 s.

From blocks to cities: Morphology structure rooted in 3D patterns and forming clusters at the block level

This article proposes a 35D form index system to quantitatively describe the 3D form of urban blocks. Utilizing the T-distributed stochastic neighbor (TSNE) embedding algorithm for cluster analysis, the visually complex and disordered urban 3D texture is translated into distinct form clusters, enabling the recognition of the overall urban form structure from the block perspective. The research methodology includes experiments conducted in the central area of Nanjing and comparative analysis in three neighboring cities: Shanghai, Hangzhou, and Suzhou. Results demonstrate the efficacy of form parameters and cluster analysis in achieving sound recognition. The four cities differ remarkably in the number and distribution structure of clusters. Shanghai has the fewest types of clusters with a compact distribution, Suzhou has the most types with a dispersed distribution, and Hangzhou and Nanjing exhibit similar characteristics, located between Shanghai and Suzhou. Correlation analysis reveals a negative relationship between the number of cluster types and the level of urban socioeconomic development in similar areas. This research implies that governments and urban planners can exploit neighborhood morphological types to devise customized spatial management and renewal strategies. The overall urban structure can be improved by strategically minimizing the quantity and distribution of neighborhood morphological types, fostering socioeconomic development.

Disruption prediction on EAST with different wall conditions based on a multi-scale deep hybrid neural network

Plasma disruption is a very dangerous event for future tokamaks and fusion reactors. Therefore, predicting disruption is crucial for ensuring the safety and performance of reactors. In this study, the features of two deep learning algorithms are integrated to establish a multi-scale hybrid network disruption predictor. Firstly, 43 diagnostic signals are extracted by a convolutional neural network (CNN), and the time information is learned by a long short-term memory network. The predictor is trained and tested on a database containing non-all-metal wall discharges. Its area under the receiver operator characteristic curve (AUC), which is a common performance metric for deep learning algorithms, reaches 0.97, and the true positive rate is , while the false positive rate is . Since EAST was upgraded in 2020, the wall condition has been upgraded from non-all-metal to all-metal. To examine the robustness of the predictor, the EAST disruption predictor is migrated to the all-metal wall experiment for the first time. It is again tested under the all-metal wall experimental data, and its warning performance decreases significantly, with an AUC of only 0.79. To improve the robustness and sensitivity of the predictor against disruptions, the convolutional attention mechanism is introduced into the CNN. After training and testing with the same data set, the warning performance for the all-metal wall data is improved, with the AUC value increasing to 0.84. To further improve the robustness of the predictor, the t-distributed stochastic neighbor embedding algorithm is employed to explore the difference of the sample distribution before and after EAST upgradation, and the transfer learning algorithm is used to reduce this difference. By applying transfer learning to a small amount of discharges from the all-metal wall data, the warning performance in the all-metal wall experiment is further improved, and the AUC value increases to 0.93. Overall, the proposed data-driven predictor is promising for applications in future devices such as ITER.

Identification of Robust Antibiotic Subgroups by Integrating Multi-Species Drug-Drug Interactions.

Previous studies have shown that antibiotics can be divided into groups, and drug-drug interactions (DDI) depend on their groups. However, these studies focused on a specific bacteria strain (i.e., Escherichia coli BW25113). Existing datasets often contain noise. Noisy labeled data may have a bad effect on the clustering results. To address this problem, we developed a multi-source information fusion method for integrating DDI information from multiple bacterial strains. Specifically, we calculated drug similarities based on the DDI network of each bacterial strain and then fused these drug similarity matrices to obtain a new fused similarity matrix. The fused similarity matrix was combined with the T-distributed stochastic neighbor embedding algorithm, and hierarchical clustering algorithm can effectively identify antibiotic subgroups. These antibiotic subgroups are strongly correlated with known antibiotic classifications, and group-group interactions are almost monochromatic. In summary, our method provides a promising framework for understanding the mechanism of action of antibiotics and exploring multi-species group-group interactions.

A Preprocessing Manifold Learning Strategy Based on t-Distributed Stochastic Neighbor Embedding.

In machine learning and data analysis, dimensionality reduction and high-dimensional data visualization can be accomplished by manifold learning using a t-Distributed Stochastic Neighbor Embedding (t-SNE) algorithm. We significantly improve this manifold learning scheme by introducing a preprocessing strategy for the t-SNE algorithm. In our preprocessing, we exploit Laplacian eigenmaps to reduce the high-dimensional data first, which can aggregate each data cluster and reduce the Kullback-Leibler divergence (KLD) remarkably. Moreover, the k-nearest-neighbor (KNN) algorithm is also involved in our preprocessing to enhance the visualization performance and reduce the computation and space complexity. We compare the performance of our strategy with that of the standard t-SNE on the MNIST dataset. The experiment results show that our strategy exhibits a stronger ability to separate different clusters as well as keep data of the same kind much closer to each other. Moreover, the KLD can be reduced by about 30% at the cost of increasing the complexity in terms of runtime by only 1-2%.

Identification of Soybean Mutant Lines Based on Dual-Branch CNN Model Fusion Framework Utilizing Images from Different Organs.

The accurate identification and classification of soybean mutant lines is essential for developing new plant varieties through mutation breeding. However, most existing studies have focused on the classification of soybean varieties. Distinguishing mutant lines solely by their seeds can be challenging due to their high genetic similarities. Therefore, in this paper, we designed a dual-branch convolutional neural network (CNN) composed of two identical single CNNs to fuse the image features of pods and seeds together to solve the soybean mutant line classification problem. Four single CNNs (AlexNet, GoogLeNet, ResNet18, and ResNet50) were used to extract features, and the output features were fused and input into the classifier for classification. The results demonstrate that dual-branch CNNs outperform single CNNs, with the dual-ResNet50 fusion framework achieving a 90.22 ± 0.19% classification rate. We also identified the most similar mutant lines and genetic relationships between certain soybean lines using a clustering tree and t-distributed stochastic neighbor embedding algorithm. Our study represents one of the primary efforts to combine various organs for the identification of soybean mutant lines. The findings of this investigation provide a new path to select potential lines for soybean mutation breeding and signify a meaningful advancement in the propagation of soybean mutant line recognition technology.

Effective Early Detection of Epileptic Seizures through EEG Signals Using Classification Algorithms Based on t-Distributed Stochastic Neighbor Embedding and K-Means.

Epilepsy is a neurological disorder in the activity of brain cells that leads to seizures. An electroencephalogram (EEG) can detect seizures as it contains physiological information of the neural activity of the brain. However, visual examination of EEG by experts is time consuming, and their diagnoses may even contradict each other. Thus, an automated computer-aided diagnosis for EEG diagnostics is necessary. Therefore, this paper proposes an effective approach for the early detection of epilepsy. The proposed approach involves the extraction of important features and classification. First, signal components are decomposed to extract the features via the discrete wavelet transform (DWT) method. Principal component analysis (PCA) and the t-distributed stochastic neighbor embedding (t-SNE) algorithm were applied to reduce the dimensions and focus on the most important features. Subsequently, K-means clustering + PCA and K-means clustering + t-SNE were used to divide the dataset into subgroups to reduce the dimensions and focus on the most important representative features of epilepsy. The features extracted from these steps were fed to extreme gradient boosting, K-nearest neighbors (K-NN), decision tree (DT), random forest (RF) and multilayer perceptron (MLP) classifiers. The experimental results demonstrated that the proposed approach provides superior results to those of existing studies. During the testing phase, the RF classifier with DWT and PCA achieved an accuracy of 97.96%, precision of 99.1%, recall of 94.41% and F1 score of 97.41%. Moreover, the RF classifier with DWT and t-SNE attained an accuracy of 98.09%, precision of 99.1%, recall of 93.9% and F1 score of 96.21%. In comparison, the MLP classifier with PCA + K-means reached an accuracy of 98.98%, precision of 99.16%, recall of 95.69% and F1 score of 97.4%.

Integrating multi-source drug information to cluster drug-drug interaction network

Characterizing drug-drug interactions is important to improve efficacy and/or slow down the evolution of antimicrobial resistance. Experimental methods are both time-consuming and laborious for characterizing drug-drug interactions. In recent years, many computational methods have been proposed to explore drug-drug interactions. However, these methods failed to effectively integrate multi-source drug information. In this study, we propose a similarity matrix fusion (SMF) method to integrate four drug information (i.e., structural similarity, pharmaceutical similarity, phenotypic similarity and therapeutic similarity). SMF combined with t-distributed stochastic neighbor embedding (t-SNE) and hierarchical clustering algorithm can effectively identify drug groups and group-group interactions are almost monochromatic (purely synergetic or purely antagonistic). To evaluate clustering quality (i.e., monochromaticity), two measures (edge purity and edge normalized mutual information) are proposed, and SMF showed the best performance. In addition, clustered drug-drug interaction network can also be used to predict new drug-drug interactions (accuracy = 0.741). Overall, SMF provides a comprehensive view to understand drug groups and group-group interactions.

Research of surface oxidation defects in copper alloy wire arc additive manufacturing based on time-frequency analysis and deep learning method

In this paper a novel method was proposed for monitoring the defects of surface oxidation in copper alloy wire arc additive manufacturing (WAAM) which using voltage sensor. Due to the good fluidity of liquid copper alloy, the dimension of copper alloy WAAM process is difficult to control especially in the height direction. This will lead to error accumulation with the layers increasing during WAAM period. The phenomenon of error accumulation will cause continuous changes in the distance between the welding wire and substrate, result in inappropriate gas protection which will lead to surface oxidation defects happen. To monitoring this defect，time-frequency analysis method was used to compare the voltage signals between normal and anomaly WAAM process. The result show that the voltage signal of abnormal weld bead appears fluctuations in its peak value, and the voltage waveform of the normal WAAM process is relatively smooth. Furthermore, continuous wavelet transform (CWT) method is used to convert one-dimensional time-series voltage signals into two-dimensional time-frequency images, which is combined with deep learning method to establish an online monitoring model to monitor the surface oxidation defects in the process of copper alloy WAAM. The accuracy of the proposed model can reach 95.83%, and t-SNE(t-distributed stochastic neighbor embedding) algorithm is used to visualize the performance of the model. The result indicates that the voltage signal characteristics of normal and abnormal WAAM processes were significantly distinguished.

A comprehensive study of clinicopathological and genetic features of neuronal intranuclear inclusion disease.

The discovery of skin intranuclear inclusions and GGC repeat expansion of NOTCH2NLC has greatly promoted the diagnosis of neuronal intranuclear inclusion disease (NIID). With highly heterogeneous clinical manifestations, NIID patients tend to be underdiagnosed at early stages. This study comprehensively studied clinical manifestations, magnetic resonance imaging (MRI), and peripheral nerve conduction in 24 NIID and 166 other neurodegenerative disease (ND) subjects. The nomogram was plotted using the "rms" package, and the t-distributed stochastic neighbor embedding algorithm was performed. Associations between skin intranuclear inclusions and NOTCH2NLC GGC repeats were further analyzed. The clinical, MRI, and peripheral nerve conduction features seriously overlapped in NIID and ND patients; they were assigned variables according to their frequency and specificity in NIID patients. A nomogram that could distinguish NIID from ND was constructed according to the assigned variables and cutoff values of the above features. The occurrence of skin intranuclear inclusions and NOTCH2NLC GGC repeats ≥ 60 showed 100% consistency, and intranuclear inclusion frequency positively correlated with NOTCH2NLC GGC repeats. A hierarchical diagnostic flowchart for definite NIID was further established. We provide a novel nomogram with the potential to realize early identification and update the diagnostic flowchart for definitive diagnosis. Moreover, this is the first study to define the association between skin pathology and NOTCH2NLC genetics in NIID.

Vascular features around the optic disc in familial exudative vitreoretinopathy: findings and their relationship to disease severity

BackgroundFamilial exudative vitreoretinopathy (FEVR) is a rare congenital disorder of retinal vascular development. We aimed to study the vascular characteristics around the optic disc in neonates with FEVR and the relationship with disease severity.MethodsA retrospective, case-control study including 43 (58 eyes) newborn patients with FEVR at stages 1 to 3 and 30 (53 eyes) age-matched normal full-term newborns was conducted. The peripapillary vessel tortuosity (VT), vessel width (VW) and vessel density (VD) were quantified by computer technology. The t-distributed stochastic neighbor embedding (t-SNE) algorithm was used to visualize the relationship between the severity of FEVR and the characteristics of perioptic disc vascular parameters.ResultsThe peripapillary VT, VW and VD were significantly increased in the FEVR group compared with the control group (P < 0.05). Subgroup analysis showed that VW and VD increased significantly with progressing FEVR stage (P < 0.05). And only VT in stage 3 FEVR was significantly increased compared with stage 1 and stage 2 (P < 0.05). After controlling the confounders, ordinal logistic regression analysis indicated that the VW (aOR: 1.75, P = 0.0002) and VD (aOR: 2.41, P = 0.0170) were significantly independent correlated with the FEVR stage, but VT (aOR: 1.07, P = 0.5454) was not correlated with FEVR staging. Visual analysis based on the t-SNE algorithm showed that peri-optic disc vascular parameters had a continuity along the direction of FEVR severity.ConclusionsIn the neonatal population, there were significant differences in peripapillary vascular parameters between patients with FEVR and normal subjects. Quantitative measurement of vascular parameters around the optic disc can be used as one of the indicators to assess the severity of FEVR.

A two-stage sorting method combining static and dynamic characteristics for retired lithium-ion battery echelon utilization

The battery echelon utilization is to sort and reuse the retired lithium-ion batteries with poor consistency, which puts forward higher requirements on how to guarantee their comprehensive consistencies after sorting. To address this issue, we combine static and dynamic characteristics as discharge capacity, temperature rise and voltage curves, and propose a two-stage sorting method for retired batteries. In the first stage, taking the discharge capacity and the temperature rise as initial sorting characteristics, the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm is used for preliminary sorting that eliminates the abnormal batteries and obtains the possible cluster number K. In the second stage, the cluster number K leads to the reference range of the initial value of K-means++ algorithm, and the dynamic features extracted from voltage curves by t-distributed Stochastic Neighbor Embedding (t-SNE) algorithm are as the input of it. The class labels of retired batteries are obtained after clustering calculation. Lastly, the validity and accuracy of the proposed method are verified based on NASA public data set, and four typical clustering evaluation indexes are applied to show its advantages. The retired batteries within the same class not only have remarkable comprehensive consistency in the three selected characteristics, but also have the consistent aging degree so that the service life of retired batteries can be prolonged to a great extent.

Identifying lightning structures via machine learning

Lightning is a fascinating yet insufficiently understood phenomenon. Very high frequency (VHF, 30–300 MHz) observations of lightning yield an ever-growing amount of data. In particular, LOFAR (LOw Frequency ARray) can reach meter and nanosecond precision with up to a million radio source locations per second. This lightning data is extremely complex, as a single lightning flash can contain hundreds of lightning channels and a myriad of different phenomena. However, so far this process has been mostly analyzed by-eye, which is very time-consuming. Thus, this increase in complexity of VHF lightning data calls for the application of machine learning algorithms. To identify structures from numerous spatio-temporal points in a high dimensional space, we designed an analysis pipeline combining a t-distributed stochastic neighbor embedding (t-SNE) algorithm and a clustering algorithm. We show that this combination allows for distinguishing correlated structures in an unsupervised approach. This novel method is a powerful tool to search vast multidimensional data sets for unique structures.

Insight into the Electronic Properties of Semiconductor Heterostructure Based on Machine Learning and First-Principles.

A first-principles approach is a powerful means of gaining insight into the intrinsic structure and properties of materials. However, with the implementation of material genetic engineering, it is still a challenging road to discover materials with high satisfaction. One alternative is to employ machine-learning techniques to mine data and predict performance. In this present contribution, the method is taken to predict the band gap opening value of graphene in a heterostructure. First, the data of 2076 binary compounds in the Materials Project library are used to achieve visual dimensionality reduction of the data set through a t-distributed stochastic neighbor embedding (t-SNE) algorithm in unsupervised learning. Then, a series of semiconductor components are screened out and form heterostructures with graphene. Second, by means of the ensemble learning EXtreme Gradient Boost (XGBoost) algorithm and support vector machine (SVM) technology, two prediction frameworks are built to predict the band gap opening value of the graphene in the system. Finally, density functional theory (DFT) is used to calculate the energy band and density of states for comparison. Analysis shows that the prediction model has an accuracy rate of 88.3%, and there is little difference between prediction results and calculation results. We anticipate that this framework model would have fascinating applications in predicting the electronic properties of various multiphase materials.