A Two-Layer Encoding Learning Swarm Optimizer Based on Frequent Itemsets for Sparse Large-Scale Multi-Objective Optimization

Traditional large-scale multi-objective optimization algorithms(LSMOEAs)encounter difficulties when dealing with sparse large-scale multi-objective optimization problems(SLM-OPs)where most decision variables are zero.As a result,many algorithms use a two-layer encoding approach to optimize binary variable Mask and real variable Dec separately.Nevertheless,existing optimizers often focus on locating non-zero variable posi-tions to optimize the binary variables Mask.However,approxi-mating the sparse distribution of real Pareto optimal solutions does not necessarily mean that the objective function is optimized.In data mining,it is common to mine frequent itemsets appear-ing together in a dataset to reveal the correlation between data.Inspired by this,we propose a novel two-layer encoding learning swarm optimizer based on frequent itemsets(TELSO)to address these SLMOPs.TELSO mined the frequent terms of multiple particles with better target values to find mask combinations that can obtain better objective values for fast convergence.Experi-mental results on five real-world problems and eight benchmark sets demonstrate that TELSO outperforms existing state-of-the-art sparse large-scale multi-objective evolutionary algorithms(SLMOEAs)in terms of performance and convergence speed.

Enhancing Evolutionary Algorithms With Pattern Mining for Sparse Large-Scale Multi-Objective Optimization Problems

Sparse large-scale multi-objective optimization problems(SLMOPs)are common in science and engineering.However,the large-scale problem represents the high dimensionality of the decision space,requiring algorithms to traverse vast expanse with limited computational resources.Furthermore,in the context of sparse,most variables in Pareto optimal solutions are zero,making it difficult for algorithms to identify non-zero variables efficiently.This paper is dedicated to addressing the challenges posed by SLMOPs.To start,we introduce innovative objective functions customized to mine maximum and minimum candidate sets.This substantial enhancement dramatically improves the efficacy of frequent pattern mining.In this way,selecting candidate sets is no longer based on the quantity of nonzero variables they contain but on a higher proportion of nonzero variables within specific dimensions.Additionally,we unveil a novel approach to association rule mining,which delves into the intricate relationships between non-zero variables.This novel methodology aids in identifying sparse distributions that can potentially expedite reductions in the objective function value.We extensively tested our algorithm across eight benchmark problems and four real-world SLMOPs.The results demonstrate that our approach achieves competitive solutions across various challenges.

Research on the Encapsulation Device for Lunar Samples

The encapsulation of lunar samples is a core research area in the third phase of the Chinese Lunar Exploration Program.The seal assembly,opening and closing mechanism(OCM),and locking mechanism are the core components of the encapsulation device of the lunar samples,and the requirements of a tight seal,lightweight,and low power make the design of these core components difficult.In this study,a combined sealing assembly,OCM,and locking mechanism were investigated for the device.The sealing architecture consists of rubber and an Ag-In alloy,and a theory was built to analyze the seal.Experiments of the electroplate Au coating on the knife-edge revealed that the hermetic seal can be significantly improved.The driving principle for coaxial double-helical pairs was investigated and used to design the OCM.Moreover,a locking mechanism was created using an electric initiating explosive device with orifice damping.By optimizing the design,the output parameters were adjusted to meet the requirements of the lunar explorer.The experimental results showed that the helium leak rate of the test pieces were not more than 5×10^(-11) Pa·m^(3)·s^(-1),the minimum power of the OCM was 0.3 W,and the total weight of the principle prototype was 2.9 kg.The explosive driven locking mechanism has low impact.This investigation solved the difficulties in achieving tight seal,light weight,and low power for the lunar explorer,and the results can also be used to explore other extraterrestrial objects in the future.

Effectiveness of Histopathological Examination of Ultrasound-guided Puncture Biopsy Samples for Diagnosis of Extrapulmonary Tuberculosis

Objective To evaluate the diagnostic value of histopathological examination of ultrasound-guided puncture biopsy samples in extrapulmonary tuberculosis(EPTB).Methods This study was conducted at the Shanghai Public Health Clinical Center.A total of 115patients underwent ultrasound-guided puncture biopsy,followed by MGIT 960 culture(culture),smear,Gene Xpert MTB/RIF(Xpert),and histopathological examination.These assays were performed to evaluate their effectiveness in diagnosing EPTB in comparison to two different diagnostic criteria:liquid culture and composite reference standard(CRS).Results When CRS was used as the reference standard,the sensitivity and specificity of culture,smear,Xpert,and histopathological examination were(44.83%,89.29%),(51.72%,89.29%),(70.11%,96.43%),and(85.06%,82.14%),respectively.Based on liquid culture tests,the sensitivity and specificity of smear,Xpert,and pathological examination were(66.67%,72.60%),(83.33%,63.01%),and(92.86%,45.21%),respectively.Histopathological examination showed the highest sensitivity but lowest specificity.Further,we found that the combination of Xpert and histopathological examination showed a sensitivity of 90.80%and a specificity of 89.29%.Conclusion Ultrasound-guided puncture sampling is safe and effective for the diagnosis of EPTB.Compared with culture,smear,and Xpert,histopathological examination showed higher sensitivity but lower specificity.The combination of histopathology with Xpert showed the best performance characteristics.

Assessment of Wet Season Precipitation in the Central United States by the Regional Climate Simulation of the WRFG Member in NARCCAP and Its Relationship with Large-Scale Circulation Biases

Assessment of past-climate simulations of regional climate models(RCMs)is important for understanding the reliability of RCMs when used to project future regional climate.Here,we assess the performance and discuss possible causes of biases in a WRF-based RCM with a grid spacing of 50 km,named WRFG,from the North American Regional Climate Change Assessment Program(NARCCAP)in simulating wet season precipitation over the Central United States for a period when observational data are available.The RCM reproduces key features of the precipitation distribution characteristics during late spring to early summer,although it tends to underestimate the magnitude of precipitation.This dry bias is partially due to the model’s lack of skill in simulating nocturnal precipitation related to the lack of eastward propagating convective systems in the simulation.Inaccuracy in reproducing large-scale circulation and environmental conditions is another contributing factor.The too weak simulated pressure gradient between the Rocky Mountains and the Gulf of Mexico results in weaker southerly winds in between,leading to a reduction of warm moist air transport from the Gulf to the Central Great Plains.The simulated low-level horizontal convergence fields are less favorable for upward motion than in the NARR and hence,for the development of moist convection as well.Therefore,a careful examination of an RCM’s deficiencies and the identification of the source of errors are important when using the RCM to project precipitation changes in future climate scenarios.

A semantic vector map-based approach for aircraft positioning in GNSS/GPS denied large-scale environment

Accurate positioning is one of the essential requirements for numerous applications of remote sensing data,especially in the event of a noisy or unreliable satellite signal.Toward this end,we present a novel framework for aircraft geo-localization in a large range that only requires a downward-facing monocular camera,an altimeter,a compass,and an open-source Vector Map(VMAP).The algorithm combines the matching and particle filter methods.Shape vector and correlation between two building contour vectors are defined,and a coarse-to-fine building vector matching(CFBVM)method is proposed in the matching stage,for which the original matching results are described by the Gaussian mixture model(GMM).Subsequently,an improved resampling strategy is designed to reduce computing expenses with a huge number of initial particles,and a credibility indicator is designed to avoid location mistakes in the particle filter stage.An experimental evaluation of the approach based on flight data is provided.On a flight at a height of 0.2 km over a flight distance of 2 km,the aircraft is geo-localized in a reference map of 11,025 km~2using 0.09 km~2aerial images without any prior information.The absolute localization error is less than 10 m.

Online identification and extraction method of regional large-scale adjustable load-aggregation characteristics

This article introduces the concept of load aggregation,which involves a comprehensive analysis of loads to acquire their external characteristics for the purpose of modeling and analyzing power systems.The online identification method is a computer-involved approach for data collection,processing,and system identification,commonly used for adaptive control and prediction.This paper proposes a method for dynamically aggregating large-scale adjustable loads to support high proportions of new energy integration,aiming to study the aggregation characteristics of regional large-scale adjustable loads using online identification techniques and feature extraction methods.The experiment selected 300 central air conditioners as the research subject and analyzed their regulation characteristics,economic efficiency,and comfort.The experimental results show that as the adjustment time of the air conditioner increases from 5 minutes to 35 minutes,the stable adjustment quantity during the adjustment period decreases from 28.46 to 3.57,indicating that air conditioning loads can be controlled over a long period and have better adjustment effects in the short term.Overall,the experimental results of this paper demonstrate that analyzing the aggregation characteristics of regional large-scale adjustable loads using online identification techniques and feature extraction algorithms is effective.

Large-Scale Multi-Objective Optimization Algorithm Based on Weighted Overlapping Grouping of Decision Variables

The large-scale multi-objective optimization algorithm(LSMOA),based on the grouping of decision variables,is an advanced method for handling high-dimensional decision variables.However,in practical problems,the interaction among decision variables is intricate,leading to large group sizes and suboptimal optimization effects;hence a large-scale multi-objective optimization algorithm based on weighted overlapping grouping of decision variables(MOEAWOD)is proposed in this paper.Initially,the decision variables are perturbed and categorized into convergence and diversity variables;subsequently,the convergence variables are subdivided into groups based on the interactions among different decision variables.If the size of a group surpasses the set threshold,that group undergoes a process of weighting and overlapping grouping.Specifically,the interaction strength is evaluated based on the interaction frequency and number of objectives among various decision variables.The decision variable with the highest interaction in the group is identified and disregarded,and the remaining variables are then reclassified into subgroups.Finally,the decision variable with the strongest interaction is added to each subgroup.MOEAWOD minimizes the interactivity between different groups and maximizes the interactivity of decision variables within groups,which contributed to the optimized direction of convergence and diversity exploration with different groups.MOEAWOD was subjected to testing on 18 benchmark large-scale optimization problems,and the experimental results demonstrate the effectiveness of our methods.Compared with the other algorithms,our method is still at an advantage.

A Large-Scale Group Decision Making Model Based on Trust Relationship and Social Network Updating

With the development of big data and social computing,large-scale group decisionmaking(LGDM)is nowmerging with social networks.Using social network analysis(SNA),this study proposes an LGDM consensus model that considers the trust relationship among decisionmakers(DMs).In the process of consensusmeasurement:the social network is constructed according to the social relationship among DMs,and the Louvain method is introduced to classify social networks to form subgroups.In this study,the weights of each decision maker and each subgroup are computed by comprehensive network weights and trust weights.In the process of consensus improvement:A feedback mechanism with four identification and two direction rules is designed to guide the consensus of the improvement process.Based on the trust relationship among DMs,the preferences are modified,and the corresponding social network is updated to accelerate the consensus.Compared with the previous research,the proposedmodel not only allows the subgroups to be reconstructed and updated during the adjustment process,but also improves the accuracy of the adjustment by the feedbackmechanism.Finally,an example analysis is conducted to verify the effectiveness and flexibility of the proposed method.Moreover,compared with previous studies,the superiority of the proposed method in solving the LGDM problem is highlighted.

Exploring device physics of perovskite solar cell via machine learning with limited samples

Perovskite solar cells(PsCs)have developed tremendously over the past decade.However,the key factors influencing the power conversion efficiency(PCE)of PSCs remain incompletely understood,due to the complexity and coupling of these structural and compositional parameters.In this research,we demon-strate an effective approach to optimize PSCs performance via machine learning(ML).To address chal-lenges posed by limited samples,we propose a feature mask(FM)method,which augments training samples through feature transformation rather than synthetic data.Using this approach,squeeze-and-excitation residual network(SEResNet)model achieves an accuracy with a root-mean-square-error(RMSE)of 0.833%and a Pearson's correlation coefficient(r)of 0.980.Furthermore,we employ the permu-tation importance(PI)algorithm to investigate key features for PCE.Subsequently,we predict PCE through high-throughput screenings,in which we study the relationship between PCE and chemical com-positions.After that,we conduct experiments to validate the consistency between predicted results by ML and experimental results.In this work,ML demonstrates the capability to predict device performance,extract key parameters from complex systems,and accelerate the transition from laboratory findings to commercialapplications.

A Fault Detection Method for Electric Vehicle Battery System Based on Bayesian Optimization SVDD Considering a Few Faulty Samples

Accurate and reliable fault detection is essential for the safe operation of electric vehicles.Support vector data description(SVDD)has been widely used in the field of fault detection.However,constructing the hypersphere boundary only describes the distribution of unlabeled samples,while the distribution of faulty samples cannot be effectively described and easilymisses detecting faulty data due to the imbalance of sample distribution.Meanwhile,selecting parameters is critical to the detection performance,and empirical parameterization is generally timeconsuming and laborious and may not result in finding the optimal parameters.Therefore,this paper proposes a semi-supervised data-driven method based on which the SVDD algorithm is improved and achieves excellent fault detection performance.By incorporating faulty samples into the underlying SVDD model,training deals better with the problem of missing detection of faulty samples caused by the imbalance in the distribution of abnormal samples,and the hypersphere boundary ismodified to classify the samplesmore accurately.The Bayesian Optimization NSVDD(BO-NSVDD)model was constructed to quickly and accurately optimize hyperparameter combinations.In the experiments,electric vehicle operation data with four common fault types are used to evaluate the performance with other five models,and the results show that the BO-NSVDD model presents superior detection performance for each type of fault data,especially in the imperceptible early and minor faults,which has seen very obvious advantages.Finally,the strong robustness of the proposed method is verified by adding different intensities of noise in the dataset.

Mechanical behavior of 2G NPR bolt anchored rock samples under static disturbance loading

The deep mining of coal resources is accompanied by severe environmental challenges and various potential engineering hazards.The implementation of NPR(negative Poisson's ratio)bolts are capable of controlling large deformations in the surrounding rock effectively.This paper focuses on studying the mechanical properties of the NPR bolt under static disturbance load.The deep nonlinear mechanical experimental system was used to study the mechanical behavior of rock samples with different anchored types(unanchored/PR anchored/2G NPR anchored)under static disturbance load.The whole process of rock samples was taken by high-speed camera to obtain the real-time failure characteristics under static disturbance load.At the same time,the acoustic emission signal was collected to obtain the key characteristic parameters of acoustic emission such as acoustic emission count,energy,and frequency.The deformation at the failure of the samples was calculated and analyzed by digital speckle software.The findings indicate that the failure mode of rock is influenced by different types of anchoring.The peak failure strength of 2G NPR bolt anchored rock samples exhibits an increase of 6.5%when compared to the unanchored rock samples.The cumulative count and cumulative energy of acoustic emission exhibit a decrease of 62.16%and 62.90%,respectively.The maximum deformation of bearing capacity exhibits an increase of 59.27%,while the failure time demonstrates a delay of 42.86%.The peak failure strength of the 2G NPR bolt anchored ones under static disturbance load exhibits an increase of 5.94%when compared to the rock anchored by PR(Poisson's ratio)bolt.The cumulative count and cumulative energy of acoustic emission exhibit a decrease of 47.16%and 43.86%,respectively.The maximum deformation of the bearing capacity exhibits an increase of 50.43%,and the failure time demonstrates a delay of 32%.After anchoring by 2G NPR bolt,anchoring support effectively reduces the risk of damage caused by static disturbance load.These results demonstrate that the support effect of 2G NPR bolt materials surpasses that of PR bolt.

Large-scale spatial data visualization method based on augmented reality

Background A task assigned to space exploration satellites involves detecting the physical environment within a certain space.However,space detection data are complex and abstract.These data are not conducive for researchers'visual perceptions of the evolution and interaction of events in the space environment.Methods A time-series dynamic data sampling method for large-scale space was proposed for sample detection data in space and time,and the corresponding relationships between data location features and other attribute features were established.A tone-mapping method based on statistical histogram equalization was proposed and applied to the final attribute feature data.The visualization process is optimized for rendering by merging materials,reducing the number of patches,and performing other operations.Results The results of sampling,feature extraction,and uniform visualization of the detection data of complex types,long duration spans,and uneven spatial distributions were obtained.The real-time visualization of large-scale spatial structures using augmented reality devices,particularly low-performance devices,was also investigated.Conclusions The proposed visualization system can reconstruct the three-dimensional structure of a large-scale space,express the structure and changes in the spatial environment using augmented reality,and assist in intuitively discovering spatial environmental events and evolutionary rules.

Numerical and theoretical study of large-scale failure of strata overlying sublevel caving mines with steeply dipping discontinuities

The deformation and fracture evolution mechanisms of the strata overlying mines mined using sublevel caving were studied via numerical simulations.Moreover,an expression for the normal force acting on the side face of a steeply dipping superimposed cantilever beam in the surrounding rock was deduced based on limit equilibrium theory.The results show the following:(1)surface displacement above metal mines with steeply dipping discontinuities shows significant step characteristics,and(2)the behavior of the strata as they fail exhibits superimposition characteristics.Generally,failure first occurs in certain superimposed strata slightly far from the goaf.Subsequently,with the constant downward excavation of the orebody,the superimposed strata become damaged both upwards away from and downwards toward the goaf.This process continues until the deep part of the steeply dipping superimposed strata forms a large-scale deep fracture plane that connects with the goaf.The deep fracture plane generally makes an angle of 12°-20°with the normal to the steeply dipping discontinuities.The effect of the constant outward transfer of strata movement due to the constant outward failure of the superimposed strata in the metal mines with steeply dipping discontinuities causes the scope of the strata movement in these mines to be larger than expected.The strata in the metal mines with steeply dipping discontinuities mainly show flexural toppling failure.However,the steeply dipping structural strata near the goaf mainly exhibit shear slipping failure,in which case the mechanical model used to describe them can be simplified by treating them as steeply dipping superimposed cantilever beams.By taking the steeply dipping superimposed cantilever beam that first experiences failure as the key stratum,the failure scope of the strata(and criteria for the stability of metal mines with steeply dipping discontinuities mined using sublevel caving)can be obtained via iterative computations from the key stratum,moving downward toward and upwards away from the goaf.

Financial Calculation Problems and Countermeasure Analysis of Large-Scale Engineering Construction Projects

The financial aspects of large-scale engineering construction projects profoundly influence their success.Strengthening cost control and establishing a scientific financial evaluation system can enhance the project’s economic benefits,minimize unnecessary costs,and provide decision-makers with a robust financial foundation.Additionally,implementing an effective cash flow control mechanism and conducting a comprehensive assessment of potential project risks can ensure financial stability and mitigate the risk of fund shortages.Developing a practical and feasible fundraising plan,along with stringent fund management practices,can prevent fund wastage and optimize fund utilization efficiency.These measures not only facilitate smooth project progression and improve project management efficiency but also enhance the project’s economic and social outcomes.

Some results of modeling D-D seismogenic pattern by the fracture model experiment of large-scale rock samples(I)

sing the natural limestone samples taken from the field with dimension of 500 mm×500 mm×1 000 mm, the D-D (dilatancy-diffusion) seismogeny pattern was modeled under the condition of water injection, which observes the time-space evolutionary features about the relative physics fields of the loaded samples from deformation, formation of microcracks to the occurrence of main rupture. The results of observed apparent resistivity show: ① The process of the deformation from microcrack to main rupture on the loaded rock sample could be characterized by the precursory spatial-temporal changes in the observation of apparent resistivity; ② The precursory temporal changes of observation in apparent resistivity could be divided into several stages, and its spatial distribution shows the difference in different parts of the rock sample; ③ Before the main rupture of the rock sample the obvious ″tendency anomaly′ and ′short-term anomaly″ were observed, and some of them could be likely considered as the ″impending earthquake ″anomaly precursor of apparent resistivity. The changes and distribution features of apparent resistivity show that they are intrinsically related to the dilatancy phenomenon of the loaded rock sample. Finally, this paper discusses the mechanism of resistivity change of loaded rock sample theoretically.

Gleaning insights from German energy transition and large-scale underground energy storage for China’s carbon neutrality

The global energy transition is a widespread phenomenon that requires international exchange of experiences and mutual learning.Germany’s success in its first phase of energy transition can be attributed to its adoption of smart energy technology and implementation of electricity futures and spot marketization,which enabled the achievement of multiple energy spatial–temporal complementarities and overall grid balance through energy conversion and reconversion technologies.While China can draw from Germany’s experience to inform its own energy transition efforts,its 11-fold higher annual electricity consumption requires a distinct approach.We recommend a clean energy system based on smart sector coupling(ENSYSCO)as a suitable pathway for achieving sustainable energy in China,given that renewable energy is expected to guarantee 85%of China’s energy production by 2060,requiring significant future electricity storage capacity.Nonetheless,renewable energy storage remains a significant challenge.We propose four large-scale underground energy storage methods based on ENSYSCO to address this challenge,while considering China’s national conditions.These proposals have culminated in pilot projects for large-scale underground energy storage in China,which we believe is a necessary choice for achieving carbon neutrality in China and enabling efficient and safe grid integration of renewable energy within the framework of ENSYSCO.

Modular Extremely Large-Scale Array Communication:Near-Field Modelling and Performance Analysis

This paper investigates the wireless communication with a novel architecture of antenna arrays,termed modular extremely large-scale array(XLarray),where array elements of an extremely large number/size are regularly mounted on a shared platform with both horizontally and vertically interlaced modules.Each module consists of a moderate/flexible number of array elements with the inter-element distance typically in the order of the signal wavelength,while different modules are separated by the relatively large inter-module distance for convenience of practical deployment.By accurately modelling the signal amplitudes and phases,as well as projected apertures across all modular elements,we analyse the near-field signal-to-noise ratio(SNR)performance for modular XL-array communications.Based on the non-uniform spherical wave(NUSW)modelling,the closed-form SNR expression is derived in terms of key system parameters,such as the overall modular array size,distances of adjacent modules along all dimensions,and the user's three-dimensional(3D)location.In addition,with the number of modules in different dimensions increasing infinitely,the asymptotic SNR scaling laws are revealed.Furthermore,we show that our proposed near-field modelling and performance analysis include the results for existing array architectures/modelling as special cases,e.g.,the collocated XL-array architecture,the uniform plane wave(UPW)based far-field modelling,and the modular extremely large-scale uniform linear array(XL-ULA)of onedimension.Extensive simulation results are presented to validate our findings.

Yarn Quality Prediction for Small Samples Based on AdaBoost Algorithm

In order to solve the problems of weak prediction stability and generalization ability of a neural network algorithm model in the yarn quality prediction research for small samples,a prediction model based on an AdaBoost algorithm(AdaBoost model) was established.A prediction model based on a linear regression algorithm(LR model) and a prediction model based on a multi-layer perceptron neural network algorithm(MLP model) were established for comparison.The prediction experiments of the yarn evenness and the yarn strength were implemented.Determination coefficients and prediction errors were used to evaluate the prediction accuracy of these models,and the K-fold cross validation was used to evaluate the generalization ability of these models.In the prediction experiments,the determination coefficient of the yarn evenness prediction result of the AdaBoost model is 76% and 87% higher than that of the LR model and the MLP model,respectively.The determination coefficient of the yarn strength prediction result of the AdaBoost model is slightly higher than that of the other two models.Considering that the yarn evenness dataset has a weaker linear relationship with the cotton dataset than that of the yarn strength dataset in this paper,the AdaBoost model has the best adaptability for the nonlinear dataset among the three models.In addition,the AdaBoost model shows generally better results in the cross-validation experiments and the series of prediction experiments at eight different training set sample sizes.It is proved that the AdaBoost model not only has good prediction accuracy but also has good prediction stability and generalization ability for small samples.

CFSA-Net:Efficient Large-Scale Point Cloud Semantic Segmentation Based on Cross-Fusion Self-Attention

Traditional models for semantic segmentation in point clouds primarily focus on smaller scales.However,in real-world applications,point clouds often exhibit larger scales,leading to heavy computational and memory requirements.The key to handling large-scale point clouds lies in leveraging random sampling,which offers higher computational efficiency and lower memory consumption compared to other sampling methods.Nevertheless,the use of random sampling can potentially result in the loss of crucial points during the encoding stage.To address these issues,this paper proposes cross-fusion self-attention network(CFSA-Net),a lightweight and efficient network architecture specifically designed for directly processing large-scale point clouds.At the core of this network is the incorporation of random sampling alongside a local feature extraction module based on cross-fusion self-attention(CFSA).This module effectively integrates long-range contextual dependencies between points by employing hierarchical position encoding(HPC).Furthermore,it enhances the interaction between each point’s coordinates and feature information through cross-fusion self-attention pooling,enabling the acquisition of more comprehensive geometric information.Finally,a residual optimization(RO)structure is introduced to extend the receptive field of individual points by stacking hierarchical position encoding and cross-fusion self-attention pooling,thereby reducing the impact of information loss caused by random sampling.Experimental results on the Stanford Large-Scale 3D Indoor Spaces(S3DIS),Semantic3D,and SemanticKITTI datasets demonstrate the superiority of this algorithm over advanced approaches such as RandLA-Net and KPConv.These findings underscore the excellent performance of CFSA-Net in large-scale 3D semantic segmentation.