Neighborhood Distribution Research Articles

Collaborative Obstacle Detection for Dual USVs Using MGNN-DANet with Movable Virtual Nodes and Double Attention

To reduce missed detections in LiDAR-based obstacle detection, this paper proposes a dual unmanned surface vessels (USVs) obstacle detection method using the MGNN-DANet template matching framework. Firstly, point cloud templates for each USV are created, and a clustering algorithm extracts suspected targets from the point clouds captured by a single USV. Secondly, a graph neural network model based on the movable virtual nodes is designed, introducing a neighborhood distribution uniformity metric. This model enhances the local point cloud distribution features of the templates and suspected targets through a local sampling strategy. Furthermore, a feature matching model based on double attention is developed, employing self-attention to aggregate the features of the templates and cross-attention to evaluate the similarity between suspected targets and aggregated templates, thereby identifying and locating another USV within the targets detected by each USV. Finally, the deviation between the measured and true positions of one USV is used to correct the point clouds obtained by the other USV, and obstacle positions are annotated through dual-view point cloud clustering. Experimental results show that, compared to single USV detection methods, the proposed method reduces the missed detection rate of maritime obstacles by 7.88% to 14.69%.

Spatial distribution of old neighborhoods based on heat-related health risks assessment: A case study of Changsha City, China

Assessing the heat-related health risks is crucial for promoting the sustainable development of cities, particularly in the face of extreme climates and urban human settlement governance. Heat health risk assessment serves as a foundational element within the risk governance framework, serving to mitigate heat-related morbidity and mortality rates. The revitalization of old neighborhoods, especially those situated in city centers with deficient facilities, emerges as a critical imperative within this context. This study chose Changsha, a city in Central China that is severely affected by high temperatures and has numerous old neighborhoods, as a case study. We developed a heat risk framework that incorporates city level risk, exposure, vulnerability, and the adaptation of old neighborhoods. Validated through data accessed from medical facilities visit, the framework effectively reflects the distribution of heat risks. By combining qualitative and quantitative methods, we investigated the correlation between heat health risk levels and the behaviors of residents in old neighborhoods. The results indicate that urban planners should prioritize comprehensive renovations in medium-risk neighborhoods, resident behavior management in low-risk neighborhoods, and the indoor thermal environment in high-risk neighborhoods. This framework plays an important role in the assessment of future spatial risk for old neighborhoods renewal.

Slowing the expansion of impervious surfaces: the key to promoting high-quality and sustainable regional development.

Sponge cities are disaster-resilient and sustainable infrastructure, and the emergence of impervious surfaces hinders the construction of sponge cities. In response to the United Nations' 2030 Sustainable Development Goals, the study of the spatio-temporal evolution and influencing factors of impervious surfaces provides an effective basis for the construction of sponge cities. In this paper, multi-source remote sensing images (Landsat 4-5 TM was used in 2000 and 2010, Landsat 8 OLI/TIRS was used in 2020) were used as data sources to construct a multi-feature impervious surface estimation model. By combining and refining the advantages of MISI, NDBI, and BUAI feature indices, we obtained the impervious surface cover of Nanchang City in 2000, 2010, and 2020. And its spatio-temporal evolution characteristics were analyzed by using the ESDA. The main factors of the impervious surface were analyzed by using the ordinary panel data model and the spatial durbin model. Results: (1) From 2000 to 2020, the impervious surface area of Nanchang City increased from 516.13 to 1075.12 km2. The overall impervious surface distribution in Nanchang City expressed a significant neighborhood distribution; (2) socio-economic factors had a positive role in promoting the impervious surface of Nanchang City. Among them, the correlation coefficient of the economic development index (0.2332), real estate investment (0.1518), and gross industrial output value (0.0453) were the most significant in the local areas; (3) the economic development index (0.2307), real estate investment (0.0251), and passenger volume (0.1679) stimulated the growth of impervious surfaces in adjacent areas, and the total population (-0.8074) had a buffering effect on adjacent areas. In order to promote the sustainable development of the region, it is necessary to propose corresponding measures and suggestions based on the impervious surface of Nanchang City.

Multi-Label Sampling based on Label Imbalance Rate and Neighborhood Distribution

Existing multi-label classification algorithms often assume that label distribution in the training set is balanced, but practical datasets frequently exhibit significant label imbalance. This imbalance affects the learning and generalization performance of the classifiers. To address the problem of label imbalance in multi-label classification, this paper proposes a new synthetic oversampling algorithm, named Multi-Label Synthetic Oversampling based on Label Imbalance Rate and Neighborhood Distribution (MLSIN). This algorithm synthesizes new samples by considering both the imbalance rate of labels and the distribution of samples in their neighborhood, aiming to improve the classifiers performance on minority labels. The rest of this chapter first introduces the evaluation metrics for multi-label classification effectiveness. Then it defines and computes the degree of label imbalance, describes the calculation of imbalance weights, and proposes a sample type correction penalty strategy, detailing the algorithm's process for selecting base and auxiliary samples., and validates the proposed method on public datasets and summarizes the experimental results.

Concrete Crack Skeleton Analysis: A Machine Vision Approach to Feature Extraction

The crack feature is the key to crack recognition in concrete surface crack detection systems based on machine vision. As the “essence” of crack, the crack skeleton shows good stability. Therefore, a feature extraction method based on inflection point recognition of concrete surface crack skeleton is proposed in this paper. The main processing steps include image preprocessing, crack skeleton extraction, and inflection point recognition. In extracting the crack skeleton, the skeleton is initially extracted using morphological operations. After identifying the endpoints based on the neighborhood distribution of the skeleton points, the burrs are tracked until the branching point using the endpoints as the starting point, and the skeleton burrs are judged and removed by setting a threshold with a burr removal rate of 100%. For the identification of inflection points based on chain code calculation, the algorithm is optimized by the concentration of the inflection points and the distance and angle between the inflection points to remove false inflection points. The test data show that the algorithm has good adaptability and the accuracy is higher than 90%. After obtaining the crack skeleton image with real inflection points, the structural features of the skeleton can be calculated, including the distance ratio before and after the inflection points and the angle formed between the inflection points. This lays the foundation for future research to realize the recognition of the same crack at different time points.

Urban greenness and its cooling effects are influenced by changes in drought, physiography, and socio-demographics in Los Angeles, CA

The multidecadal change in urban microclimate and greenness, particularly in response to drought and a warming climate, has implications for urban residents' well-being. Urban greenness, temperature, and vegetative cooling vary spatially. However, the dynamics of the relationships among these variables and their influencing factors are poorly characterized. Using the Los Angeles Urban Region, USA as a case study we evaluate the dynamics among urban vegetation and climate through an evaluation of satellite-based observations between 1985 and 2021. We hypothesize that microclimate changes are driven by water demand and aridity, with increasing aridity enhancing transpiration and vegetation-cooling, but that irrigation variation, assessed through proxy demographic variables of income modify water availability. Our results show that the L.A. region warmed by 0.13 °C/year, NDVI increased annually by 4.81 × 10−4, and vegetative cooling increased by 0.08 °C/NDVI/year. A consequence of these dynamics was that the luxury effect of income as a mediator of NDVI and LST declined 41% and 28%, respectively, between 1990 and 2020. The changes in urban microclimates over time and from drought are affected by social and physiographic variables associated with water availability and water demand and are increasingly leading to less racially equitable neighborhood distributions of heat.

Structure-Driven Representation Learning for Deep Clustering

As an important branch of unsupervised learning methods, clustering makes a wide contribution in the area of data mining. It is well known that capturing the group-discriminative properties of each sample for clustering is crucial. Among them, deep clustering delivers promising results due to the strong representational power of neural networks. However, most of them adopt sample-level learning strategies, and the standalone data point barely captures its holistic cluster’s context and may undergo sub-optimal cluster assignment. To tackle this issue, we propose a Structure-driven Representation Learning (SRL) method by introducing latent structure information into the representation learning process at both the local and global levels. Specifically, a local-structure-driven sample representation strategy is proposed to approximate the estimation of data distribution, which models the neighborhood distribution of samples with potential structure information and exploits statistical dependencies between them to improve cluster consistency. A global-structure-driven cluster representation strategy is designed, where the context of each cluster is sufficiently encoded according to its samples (exemplar-theory) and corresponding prototype (prototype-theory). In this case, each cluster can only be related to its most similar samples, and different clusters are separated as much as possible. These two models are seamlessly combined into a joint optimization problem, which can be efficiently solved. Experiments on six widely-used datasets demonstrate the superiority of SRL over state-of-the-art clustering methods.

Unraveling the functional dark matter through global metagenomics.

Metagenomes encode an enormous diversity of proteins, reflecting a multiplicity of functions and activities1,2. Exploration of this vast sequence space has been limited to a comparative analysis against reference microbial genomes and protein families derived from those genomes. Here, to examine the scale of yet untapped functional diversity beyond what is currently possible through the lens of reference genomes, we develop a computational approach to generate reference-free protein families from the sequence space in metagenomes. We analyse 26,931 metagenomes and identify 1.17 billion protein sequences longer than 35 amino acids with no similarity to any sequences from 102,491 reference genomes or the Pfam database3. Using massively parallel graph-based clustering, we group these proteins into 106,198 novel sequence clusters with more than 100 members, doubling the number of protein families obtained from the reference genomes clustered using the same approach. We annotate these families on the basis of their taxonomic, habitat, geographical and gene neighbourhood distributions and, where sufficient sequence diversity is available, predict protein three-dimensional models, revealing novel structures. Overall, our results uncover an enormously diverse functional space, highlighting the importance of further exploring the microbial functional dark matter.

Faster Property Testers in a Variation of the Bounded Degree Model

Property testing algorithms are highly efficient algorithms that come with probabilistic accuracy guarantees. For a property P , the goal is to distinguish inputs that have P from those that are far from having P with high probability correctly, by querying only a small number of local parts of the input. In property testing on graphs, the distance is measured by the number of edge modifications (additions or deletions) that are necessary to transform a graph into one with property P . Much research has focused on the query complexity of such algorithms, i. e., the number of queries the algorithm makes to the input, but in view of applications, the running time of the algorithm is equally relevant. In (Adler, Harwath, STACS 2018), a natural extension of the bounded degree graph model of property testing to relational databases of bounded degree was introduced, and it was shown that on databases of bounded degree and bounded tree-width, every property that is expressible in monadic second-order logic with counting (CMSO) is testable with constant query complexity and sublinear running time. It remains open whether this can be improved to constant running time. In this article we introduce a new model, which is based on the bounded degree model, but the distance measure allows both edge (tuple) modifications and vertex (element) modifications. We show that every property that is testable in the classical model is testable in our model with the same query complexity and running time, but the converse is not true. Our main theorem shows that on databases of bounded degree and bounded tree-width, every property that is expressible in CMSO is testable with constant query complexity and constant running time in the new model. Our proof methods include the semilinearity of the neighborhood histograms of databases having the property and a result by Alon (Proposition 19.10 in Lovász, Large networks and graph limits, 2012) that states that for every bounded degree graph \(\mathcal {G}\) there exists a constant size graph \(\mathcal {H}\) that has a similar neighborhood distribution to \(\mathcal {G}\) . It can be derived from a result in (Benjamini et al., Advances in Mathematics 2010) that hyperfinite hereditary properties are testable with constant query complexity and constant running time in the classical model (and hence in the new model). Using our methods, we give an alternative proof that hyperfinite hereditary properties are testable with constant query complexity and constant running time in the new model. We argue that our model is natural and our meta-theorem showing constant-time CMSO testability supports this.

Fault Diagnosis Method Based on CND-SMOTE and BA-SVM Algorithm

The problem of unbalanced data classification has gotten extensive attention in the past few years. Unbalanced sample data makes the fault diagnosis and classification accuracy rate low, and the capability to classify minority-class fault samples is restricted. To address the problem that the classification algorithm in machine learning has the insufficient capability to identify minority class samples for unbalanced sample data classification problems. Therefore, this paper proposes an improved support vector machine (SVM) classification method based on the synthetic minority over-sampling technique (SMOTE). For the sampler, an improved synthetic minority over-sampling technique based on the characteristics of neighborhood distribution (CND-SMOTE) algorithm is used to equilibrate the minority class samples and the majority class samples. For the classifier, the parameter optimization method of support vector machines based on the bat algorithm (BA-SVM) is used to solve the multi-classification problem of faulty samples. Finally, experimental results prove that the CND-SMOTE+BA-SVM algorithm can synthesize high-quality minority fault samples, increase the classification accuracy rate of fault samples, and decrease the time spent on the classification.

Toward a Healthy Urban Living Environment: Assessing 15-Minute Green-Blue Space Accessibility

Exposure to green-blue space has been shown to be associated with better physical and mental health outcomes. The advent of COVID-19 has underlined the importance for people to have access to green-blue spaces in proximity to their residences due to pandemic-related restrictions on activity space. The implementation of the 15-min concept, which advocates that people should be able to reach locations of essential functions like green-blue spaces within 15 min of active travel, can bring green-blue spaces nearer to where people live. Nonetheless, there is still a lack of understanding of the social and spatial (in)equality in 15-min green-blue space accessibility by active travel in cities seeking to embrace the concept, such as Hong Kong. This study explores 15-min green-blue space accessibility by walking and cycling in Hong Kong to reveal the distribution of disadvantaged neighborhoods. The results show that neighborhoods in Kowloon’s districts are the most disadvantaged in accessing green-blue spaces within 15 min of active travel. Our study provides policymakers with valuable insights and knowledge conducive to formulating policies aimed at reducing inequality in 15-min accessibility.

Neighborhood distribution of availability of newer tobacco products: A US four-site study, 2021

Audits of tobacco retailers can identify marketing patterns as newer tobacco products are introduced in the US. Our study examined store and neighborhood correlates of availability of nicotine pouches and disposable e-cigarettes in four US sites. We conducted standardized store audits of n = 242 tobacco retailers in 2021 in different states: New Jersey, Kentucky, North Carolina, and New York. We geocoded stores linking them with census tract demographics. We conducted unadjusted and adjusted Poisson regression of availability of each product with correlates of the proportion of Non-Hispanic White residents, households under poverty, proximity to schools, site, and store type. Nicotine pouches and disposable e-cigarettes were each available in around half the stores overall, but availability differed across sites (range: 76 %–32 %). In adjusted analyses, nicotine pouches were less likely to be available in each store type vs chain convenience (IRR range 0.2–0.6) and more likely in stores in census tracts with a greater percentage of non-Hispanic White residents (IRR range 1.8–2.3). In contrast, disposable e-cigarettes were more likely to be available in tobacco/vape shops (IRR 1.9 (1.4–2.5) than convenience stores and less likely in non-specialty store types like groceries (IRR 0.2 (0.1–0.4). Newer tobacco products like nicotine pouches and disposable e-cigarettes were widely available in stores across sites, but retail marketing patterns appear to differ. As these product types become subject to increased regulation as they go through the FDA pre-market authorization process, understanding patterns and changes in the retail environment is critical to inform potential policies regulating their sale and marketing.

Exploring surface urban heat island (SUHI) intensity and its implications based on urban 3D neighborhood metrics: An investigation of 57 Chinese cities

Excessive urban temperature exerts a substantially negative impact on urban sustainability. Three-dimensional (3D) landscapes have a great impact on urban thermal environments, while their heat conditions and driving factors still remain unclear. This study mapped urban 3D neighborhoods and their associated SUHI (surface urban heat island) intensities in summer daytime across 57 Chinese cities, and then explored their relationships, driving factors as well as implications. Nine categories of urban 3D neighborhoods existed in Chinese cities and the 3D neighborhood of High Density & Medium Rise (HDMR) contributed the largest share of urban areas. The distribution of 3D neighborhoods varied among cities due to their distinct natural and economic traits. The average SUHI intensity can amount to 4.27 °C across all Chinese 3D neighborhoods. High Density & Low Rise (HDLR) and HDMR presented higher SUHI intensities than other 3D neighborhoods in China. Urban green space (UGI) and building height (BH) had great influences on SUHI intensities. The relative contribution of UGI decreased with the increase of building density and building height, but BH presented the opposite trend. The interaction of urban 3D landscapes and function zones led to highly complicated urban thermal environments, with higher SUHI intensities in industrial zones. Besides, the SUHI intensities of 3D neighborhoods presented great diurnal and seasonal variations, with higher SUHI intensities in HDHR and HDMR at nighttime in winter and summer. What's more, urban residents may suffer unequal heat risk inside cities due to the deviations of SUHI intensities among different 3D neighborhoods. It could be a highly effective way to mitigate SUHI effects in cities by increasing urban greening and improving urban ventilation.

Affinity estimation models of proteins for intelligent drug design based on pseudoconvolutions and nonlinear regressors

Purpose: The emergence of new viruses and, consequently, new diseases make the rapid and precise design of new drugs increasingly necessary. With the availability of large databases of proteins and affinity measures, it is possible to build scoring functions for predicting molecular affinity. These functions are fundamental to intelligent drug design. Objective: In this work, we propose a scoring function to predict affinity between two proteins. The method is based on extracting features by transfer learning on sequences represented on pseudo-convolutions. Method: The pseudo-convolutions organize the sequences into base neighborhood distributions. Each distribution is then represented by an image. Two proteins are then transformed into two images that are concatenated together, forming the third image. Through deep transfer learning, this resulting image is then represented by a vector of attributes, which have dimensionality reduced by Random Forest. Finally, the vector of attributes reduced is applied to a regression learning machine that returns the degree of affinity of the two proteins. Results: We used the Affinity Benchmark Version 2 database. 145 complexes were used for model training and 35 for testing. The results showed a performance equal to or better than the state-of-the-art methods of evaluating protein affinity, considering the correlation coefficients of Pearson, Spearman and Kendall. The best results were 0.66, 0.70, and 0.52. Conclusion: The proposed method can characterize protein sequences so that the binding affinity between two proteins can be estimated without simulating the three-dimensional structure of the complex.

Neighborhood Distribution of Unsheltered Homelessness and its Temporal Changes: Evidence from Los Angeles

Using the 2016–2020 point-in-time homeless count data, this study investigates neighborhood characteristics associated with the levels of and changes in unsheltered homeless population density in Los Angeles. The results show that unsheltered homeless people in the study area are heavily concentrated in and around the city center, and in neighborhoods with greater access to shelters and lower socioeconomic status. Notably, neighborhoods closer to the city center experienced a relatively large increase in unsheltered homelessness during the study period, implying a persistent spatial concentration of unsheltered homelessness. The results consistently indicate that residential land share, Hispanic resident share, and the number of bridges in the baseline year are significant predictors of relative changes in unsheltered homelessness in subsequent years, whereas access to shelters and poverty rates are not. This study’s findings provide several important policy implications that could potentially help prevent and mitigate unsheltered homelessness.

The spatial distribution of businesses and neighborhoods: What industries match or mismatch what neighborhoods?

Using the Longitudinal Employer-Household Dynamic (LEHD) and American Community Survey (ACS), we evaluate the spatial relationship between businesses and neighborhoods through exploring the effects of homeownership rates on the job counts of businesses located at various distances to the neighborhoods. The effects of interest are evaluated for neighborhoods of different income levels and businesses of various types. To construct the spatial framework, we first cluster the businesses and then draw residential donut rings of different radii surrounding the business clusters. To control for any fixed locational and temporal features affecting the location choices of both households and businesses, we employ a fixed-effect (FE) identification strategy, where a positive (negative) coefficient suggests that the businesses match (mismatch) the neighborhoods with a higher concentration of homeowners. Our findings show that no industries mismatch the higher-income homeowners, whereas a mismatch exists between their lower-income counterparts and the industries of food and entertainment; the industries that match homeowners of both income groups include retail trade, whole trade, and education. In addition to these three sectors, health, professional services, and entertainment match the higher-income homeowners, whereas construction and transportation match the lower-income ones; for retail trade and whole trade, the matching occurs at a shorter distance between the lower income neighborhoods and businesses. Our results can assist the government in perfecting zoning regulations, developing subsidizing policies, and constructing affordable housing programs.

IVEN: A quantitative tool to describe 3D cell position and neighbourhood reveals architectural changes in FGF4-treated preimplantation embryos.

Architectural changes at the cellular and organism level are integral and necessary to successful development and growth. During mammalian preimplantation development, cells reduce in size and the architecture of the embryo changes significantly. Such changes must be coordinated correctly to ensure continued development of the embryo and, ultimately, a successful pregnancy. However, the nature of such transformations is poorly defined during mammalian preimplantation development. In order to quantitatively describe changes in cell environment and organism architecture, we designed Internal Versus External Neighbourhood (IVEN). IVEN is a user-interactive, open-source pipeline that classifies cells into different populations based on their position and quantifies the number of neighbours of every cell within a dataset in a 3D environment. Through IVEN-driven analyses, we show how transformations in cell environment, defined here as changes in cell neighbourhood, are related to changes in embryo geometry and major developmental events during preimplantation mammalian development. Moreover, we demonstrate that modulation of the FGF pathway alters spatial relations of inner cells and neighbourhood distributions, leading to overall changes in embryo architecture. In conjunction with IVEN-driven analyses, we uncover differences in the dynamic of cell size changes over the preimplantation period and determine that cells within the mammalian embryo initiate growth phase only at the time of implantation.

Detecting overlapping communities using ensemble-based distributed neighbourhood threshold method in social networks

In this paper, we have proposed a novel overlapping community detection algorithm based on an ensemble approach with a distributed neighbourhood threshold method (EnDNTM). EnDNTM uses pre-partitioned disjoint communities generated by the ensemble mechanism and then analyzes the neighbourhood distribution of boundary nodes in disjoint communities to detect overlapping communities. It is a form of seed-based global method since boundary nodes are considered as seeds and become the starting point for detecting overlapping communities. A threshold value for each boundary node is used as the minimum influence by the neighbours of a node in order to determine its belongingness to any community. The effectiveness of the EnDNTM algorithm has been demonstrated by testing with five synthetic benchmark datasets and fifteen real-world datasets. The performance of the EnDNTM algorithm was compared with seven overlapping community detection algorithms. The F1-score, normalized mutual information ONMI and extended modularity Qo⁢v metrics were used to measure the quality of the detected communities. EnDNTM outperforms comparable algorithms on 4 out of 5 synthetic benchmarks datasets, 11 out of 15 real world datasets and gives comparable results with the remaining datasets. Experiments on various synthetic and real world datasets reveal that for a majority of datasets, the proposed ensemble-based distributed neighbourhood threshold method is able to select the best disjoint clusters produced by a disjoint method from a collection of methods for detecting overlapping communities.

Within-stand temporal and spatial dynamics of tree neighbourhood density and species composition: even-agedvssingle-tree selection forests

AbstractTree neighbourhood conditions vary greatly through time for individual trees and throughout a forest at any given time. Therefore, many trees experience very different neighbourhood conditions to the stand mean. Although changes in stand density or species composition are often recommended to mitigate the effects of global change on forest functioning, these recommendations are often based on stand means and lack guidance on acceptable variability around the means or acceptable ranges in neighbourhood conditions. This study examined how the spatial and temporal variability in tree neighbourhood conditions varies between species and with stand structural characteristics. A trees neighbourhood, defined by all trees within a 10-m radius, was quantified in terms of basal area GN, relative height rhN (tree height divided by mean height of all trees in neighbourhood), species proportion PropN and species richness RN. Weibull functions were used to describe the frequency distributions of GN, rhN, PropN and RN within the stands. Equations were developed to describe the Weibull shape, scale and location parameters as functions of the species, stand basal area, mean tree diameter and management (even-aged vs single-tree selection forests). All of these variables significantly influenced the frequency distributions of neighbourhood characteristics. These equations can be used to show the proportion of trees experiencing significantly higher or lower values of a given characteristic than the stand mean or other threshold values relevant for management objectives. Single-tree selection forests had much greater ranges in neighbourhoods. There were also large temporal changes in tree neighbourhoods, especially in single-tree selection forests. Large trees had not experienced better neighbourhood conditions than smaller trees of approximately the same age, when considered over the long-term. Given that foresters typically consider tree neighbourhoods when marking trees for thinning, and that trees respond to tree neighbourhood conditions rather than to mean stand conditions, stand-level recommendations may be easier to transfer into practice and to interpret, when accompanied by neighbourhood-level information that indicates the distribution of neighbourhoods that actually exist within a stand.

Incremental Factorization of Big Time Series Data with Blind Factor Approximation

Extracting the latent factors of big time series data is an important means to examine the dynamic complex systems under observation. These low-dimensional and “small” representations reveal the key insights to the overall mechanisms, which can otherwise be obscured by the notoriously high dimensionality and scale of big data as well as the enormously complicated interdependencies amongst data elements. However, grand challenges still remain: (1) to incrementally derive the multi-mode factors of the augmenting big data and (2) to achieve this goal under the circumstance of insufficient a priori knowledge. This study develops an incrementally parallel factorization solution (namely I-PARAFAC ) for huge augmenting tensors (multi-way arrays) consisting of three phases over a cutting-edge GPU cluster: in the “giant-step” phase , a variational Bayesian inference ( VBI ) model estimates the distribution of the close neighborhood of each factor in a high confidence level without the need for a priori knowledge of the tensor or problem domain; in the “baby-step” phase , a massively parallel Fast-HALS algorithm (namely G-HALS ) has been developed to derive the accurate subfactors of each subtensor on the basis of the initial factors; in the final fusion phase , I-PARAFAC fuses the known factors of the original tensor and those accurate subfactors of the “increment” to achieve the final full factors. Experimental results indicate that: (1) the VBI model enables a blind factor approximation , where the distribution of the close neighborhood of each final factor can be quickly derived (10 iterations for the test case). As a result, the model of a low time complexity significantly accelerates the derivation of the final accurate factors and lowers the risks of errors; (2) I-PARAFAC significantly outperforms even the latest high performance counterpart when handling augmenting tensors, e.g., the increased overhead is only proportional to the increment while the latter has to repeatedly factorize the whole tensor, and the overhead in fusing subfactors is always minimal; (3) I-PARAFAC can factorize a huge tensor (volume up to 500 TB over 50 nodes) as a whole with the capability several magnitudes higher than conventional methods, and the runtime is in the order of $\frac{1}{n}$ 1 n to the number of compute nodes; (4) I-PARAFAC supports correct factorization-based analysis of a real 4-order EEG dataset captured from a variety of epilepsy patients. Overall, it should also be noted that counterpart methods have to derive the whole tensor from the scratch if the tensor is augmented in any dimension; as a contrast, the I-PARAFAC framework only needs to incrementally compute the full factors of the huge augmented tensor.