Constrained Clustering Method Research Articles

Clustering policy for urban competitiveness: One‐size‐fits‐all? Evidence from Tehran

Urban competitiveness is commonly attributed to necessitating particular types of socio‐spatial frameworks such as clustering on the basis of models of Marshall–Arrow–Romer (MAR), Porter, or Jacobs, since proximity is known as a key perquisite for knowledge spillover, innovation, and hence urban competitiveness. Nevertheless, it is unjustifiable to unconditionally endorse the ‘clustering’ policy. Consequently, this research aims to examine the spatiality and geography of competition and innovation contingent upon varying spatial prerequisites and contextual circumstances. In this regard, two questions have been raised: under what circumstances urban competitiveness depends on spatial agglomeration, and whether this agglomeration follows a clustering development pattern. Through the implementation of a causal research design utilizing both in‐depth interviews and social survey methodology, this study has determined the innovative characteristics and techniques for knowledge acquisition, as well as spatial behaviors and preferences of competitive industries located within Tehran. The spatially constrained multivariate clustering method was employed to group competitive industries according to their distinct substantive dissimilarities, socio‐spatial behaviors, and characteristics. Then, on the basis of the structural equation modeling, the most appropriate spatial development frameworks of competitiveness were extracted with regard to the specific conditions of Tehran. Findings show that innovation, and hence competitiveness, are not always the result of clustering; clusters, as a spatial policy‐making tool, have been found to be highly conducive to collaborative and intensive knowledge production, as they provide access to spillover knowledge, thereby increasing competitiveness. Despite complex economic dependencies and collaborations, some competitive industries have greater mobility and are footloose due to their Schumpeterian linear knowledge source and virtual nature. What matters for this elusive, light, fluid category free from the shackles of place is institutional and organizational proximity, which proves its conditions are up to the government. As the main finding of the research, the geography and spatiality of innovation, and therefore the competitiveness, depend on two factors, including: (a) the relationships and interactions, that is, types of activity units and (b) the source and method of acquiring knowledge required for innovation and competitive advantage. In some activities, on the basis of their knowledge intensity, interaction type, and knowledge source, competitiveness is not related to clusters; in others, it depends on a specific location; and some are independent of location.

Spatial differentiation and geographical similarity of traditional villages--Take the Yellow River Basin and the Yangtze River Basin as examples.

The preservation and advancement of Traditional Villages are intricately linked to the perpetuation of cultural heritage. However, existing research on the spatial disparities among Traditional Villages has yet to consider the impact of cultural factors. Moreover, the geographical similarities shared by Traditional Villages have not been thoroughly examined. Therefore, this study takes the Yellow River Basin and the Yangtze River Basin, both pivotal in the genesis of Chinese culture, as case studies. We employ an Optimal Parameter-based GeoDetector alongside the Spatially Constrained Multivariate Clustering method to delve into the factors influencing the spatial differentiation and geographical similarities of Traditional Villages within these two significant river basins. Subsequently, we offer recommendations for fostering the sustainable preservation and development of Traditional Villages in these regions. The results indicate that the Rural Residents Per Capita Disposable Income has the greatest impact on the spatial differentiation of Traditional Villages in the Yellow River Basin, while the Density of National Intangible Cultural Heritage Inheritors has the most significant influence in the Yangtze River Basin. The interaction effects of the driving factors are more powerful, with a q-value of 0.9544 for the interplay between the Density of National Cultural Protection Units and the Tourism Income in the Yellow River Basin and a value of 0.9099 for the interaction between the Density of National Intangible Cultural Heritage Inheritors and the Transportation in the Yangtze River Basin. Regarding geographical similarity, the Traditional Villages in the Yellow River Basin are divided into three major clusters, while those in the Yangtze River Basin are divided into two.

FP-Age: Leveraging Face Parsing Attention for Facial Age Estimation in the Wild.

Image-based age estimation aims to predict a person's age from facial images. It is used in a variety of real-world applications. Although end-to-end deep models have achieved impressive results for age estimation on benchmark datasets, their performance in-the-wild still leaves much room for improvement due to the challenges caused by large variations in head pose, facial expressions, and occlusions. To address this issue, we propose a simple yet effective method to explicitly incorporate facial semantics into age estimation, so that the model would learn to correctly focus on the most informative facial components from unaligned facial images regardless of head pose and non-rigid deformation. To this end, we design a face parsing-based network to learn semantic information at different scales and a novel face parsing attention module to leverage these semantic features for age estimation. To evaluate our method on in-the-wild data, we also introduce a new challenging large-scale benchmark called IMDB-Clean. This dataset is created by semi-automatically cleaning the noisy IMDB-WIKI dataset using a constrained clustering method. Through comprehensive experiment on IMDB-Clean and other benchmark datasets, under both intra-dataset and cross-dataset evaluation protocols, we show that our method consistently outperforms all existing age estimation methods and achieves a new state-of-the-art performance. To the best of our knowledge, our work presents the first attempt of leveraging face parsing attention to achieve semantic-aware age estimation, which may be inspiring to other high level facial analysis tasks.

Incorporation of Spatially Heterogeneous Area Partitioning into Vector-Based Cellular Automata for Simulating Urban Land-Use Changes

In cellular automata (CA) modeling, spatial heterogeneity can be delineated by geographical area partitioning. The dual constrained space clustering method is a prevalent approach for providing an objective and effective representation of differences within urban regions. However, previous studies faced issues by ignoring spatial heterogeneity, which could lead to an over- or under-estimation of the simulation results. Accordingly, this study attempts to incorporate spatially heterogeneous area partitioning into vector-based cellular automata (VCA), producing more accurate and reliable simulations of urban land-use change. First, an area partition strategy with DSC algorithm was employed to generate multiple relatively homogeneous sub-regions, which can effectively capture the spatial heterogeneity in the distribution of land-use change factors. Second, UrbanVCA, a brand-new VCA-based framework, was utilized for simulating land-use changes in distinct urban partitions. Finally, the constructed partitioned VCA model was applied to simulate rapid urban development in Jiangyin city from 2012 to 2017. The results indicated that the combination of DSC clustering and UrbanVCA model could obtain satisfying results as the average FoM values for the partitions and the entire study area exceeded 0.22. Furthermore, a comparative analysis of results from traditional area-partitioned CA models revealed that the proposed area partitioning approach had the potential to yield more accurate simulation outcomes as the FoM values were higher and SHDI and LSI metrics were closer to real-world observations, indicating its good performance in simulating fragmented urban landscapes.

Dual Mutual Information Constraints for Discriminative Clustering

Deep clustering is a fundamental task in machine learning and data mining that aims at learning clustering-oriented feature representations. In previous studies, most of deep clustering methods follow the idea of self-supervised representation learning by maximizing the consistency of all similar instance pairs while ignoring the effect of feature redundancy on clustering performance. In this paper, to address the above issue, we design a dual mutual information constrained clustering method named DMICC which is based on deep contrastive clustering architecture, in which the dual mutual information constraints are particularly employed with solid theoretical guarantees and experimental validations. Specifically, at the feature level, we reduce the redundancy among features by minimizing the mutual information across all the dimensionalities to encourage the neural network to extract more discriminative features. At the instance level, we maximize the mutual information of the similar instance pairs to obtain more unbiased and robust representations. The dual mutual information constraints happen simultaneously and thus complement each other to jointly optimize better features that are suitable for the clustering task. We also prove that our adopted mutual information constraints are superior in feature extraction, and the proposed dual mutual information constraints are clearly bounded and thus solvable. Extensive experiments on five benchmark datasets show that our proposed approach outperforms most other clustering algorithms. The code is available at https://github.com/Li-Hyn/DMICC.

Urban morphological regionalization based on 3D building blocks—A case in the central area of Chengdu, China

Three-dimensionality (3D), the main feature of urban morphology since rapid urban development, has been a trending topic on urban morphology research. Many 3D morphological indices have been proposed to describe urban morphology in current studies. However, the understanding of those indices is independent and isolated, and the spatial relationship of these indices is rarely considered, which makes it difficult to grasp the overall morphological characteristics and its spatial structure. Based on 3D buildings in the central area of Chengdu City, nine morphological indices are constructed. Then, a spatially constrained multivariate clustering method is applied for the comprehensive regionalization of urban morphology after index correlation inspection. The results show that: (1) four morphological indices from height, density and volume can effectively describe the characteristics of urban morphology in the study area respectively; (2) similar to the landform map, a ten-region map is generated that can clearly and intuitively demonstrate 3D morphological zones of integrated characteristics; and (3) a “two-ring with multi-sector” spatial structure is further abstracted from the regional map. This study contributes to acknowledging the spatial structure of urban morphology from a vertical perspective, and the framework can be reduplicated in other cities, which are meaningful to urban management, planning, and construction.

Marine Environmental Regionalization for the Beibu Gulf Based on a Physical-Biological Model

A physical–biological ocean model was employed to investigate characteristics of the Beibu Gulf in the northwest South China Sea (SCS) from 2011 to 2015. We adopted the spatially constrained multivariate clustering method to determine the refined marine environmental regionalization using 10 variables from the model output, and compared regionalization differences in ENSO (El Niño–Southern Oscillation) years. The simulated physical and biochemical variables display a wide spectrum of patterns in space and time. The regionalization maps indicated that the Qiongzhou Strait and its adjacent area can be classified as a separate region, characterized by the rich presence of nutrients, phytoplankton, zooplankton, and detritus, owing to the water invasion from the western Guangdong estuary. As a result of the invasive progression of the SCS, the northern and southern gulf show distinct features over a boundary near 20° N. In the La Niña year (2011), the classified boundary of the Qiongzhou Strait-northeastern gulf moved southwards due to enhanced phytoplankton growth. In the El Niño year (2015), the current collision from the northern gulf and SCS resulted in the boundary of the northern and southern gulf moving to approximately 19° N. These results provide useful guidance on subregional marine management and subregional studies for the gulf.

A Novel Two-Stage Multi-Layer Constrained Spectral Clustering Strategy for Intentional Islanding of Power Grids

Intentional islanding can be considered as the last action to prevent power grids from severe blackouts. In this strategy, the endangered network is deliberately decomposed into self-sustained islands to improve the power grid resilience, reliability, and security. In this way, this paper develops a novel optimal intentional islanding solution to deal with deliberate physical attacks on the power system. The proposed solution employs a modified multi-layer constrained clustering method based on multi-layer graphs via subspace analysis on Grassmann manifolds clustering. The main objectives are to minimize the active and reactive powers disruptions crossways for decomposed islands, as well as grouping the generators with high-frequency change similarity so that stable operation of each self-supplied grid is ensured. This technique guarantees that each island is only comprised of generators that are synchronized with each other. The proposed multi-layer technique increases the stability of the island by implementing different criteria such as frequency, active power, and reactive power. To better display the improvement, the method is compared with the single layer constrained clustering that can only handle one criterion at a time. The proposed intentional islanding technique is applied to IEEE-9 bus, IEEE-39 bus, and IEEE-118 bus as small, medium, and large-scale networks, respectively.

Urban Structure, Subway Systemand Housing Price: Evidence from Beijing and Hangzhou, China

Using housing market data of Beijing and Hangzhou, China, we conduct a case study to detect how the difference of urban structure can affect the relationship between the subway system and housing prices. To quantify the characteristics of urban structure, we propose a constrained clustering method, which can not only reveal the spatial heterogeneity of the housing market, but also provides a link between heterogeneity and the underlying urban structure. Applying constrained clustering to Beijing and Hangzhou, we find that the relationship between accessibility to metro stations and housing prices is weak and vulnerable, while the improvement of commuting efficiency, measured by a key variable, the metro index, does have a robust connection to metro premium on housing units. In particular, only a large metro index can be associated with a positive metro premium. Structural features, such as the size of urban core and the existence of multiple sub-centers, influence the metro premium by affecting the value and spatial distribution of the metro index. The evidence from Beijing and Hangzhou supports that in a mono-centric city, the size of the urban core is positively associated with the metro index and the metro premium, while in a poly-centric city with a small urban core, the metro index tends to be lower in the core region and higher in the satellite regions, which enforces the metro premium to be negative in the core while positive outside of the core.

PSI: A probabilistic semantic interpretable framework for fine‐grained image ranking

Image Ranking is one of the key problems in information science research area. However, most current methods focus on increasing the performance, leaving the semantic gap problem, which refers to the learned ranking models are hard to be understood, remaining intact. Therefore, in this article, we aim at learning an interpretable ranking model to tackle the semantic gap in fine‐grained image ranking. We propose to combine attribute‐based representation and online passive‐aggressive (PA) learning based ranking models to achieve this goal. Besides, considering the highly localized instances in fine‐grained image ranking, we introduce a supervised constrained clustering method to gather class‐balanced training instances for local PA‐based models, and incorporate the learned local models into a unified probabilistic framework. Extensive experiments on the benchmark demonstrate that the proposed framework outperforms state‐of‐the‐art methods in terms of accuracy and speed.

Constrained clustering with a complex cluster structure

In this contribution we present a novel constrained clustering method, Constrained clustering with a complex cluster structure (C4s), which incorporates equivalence constraints, both positive and negative, as the background information. C4s is capable of discovering groups of arbitrary structure, e.g. with multi-modal distribution, since at the initial stage the equivalence classes of elements generated by the positive constraints are split into smaller parts. This provides a detailed description of elements, which are in positive equivalence relation. In order to enable an automatic detection of the number of groups, the cross-entropy clustering is applied for each partitioning process. Experiments show that the proposed method achieves significantly better results than previous constrained clustering approaches. The advantage of our algorithm increases when we are focusing on finding partitions with complex structure of clusters.

Constrained clustering method for class-based storage location assignment in warehouse

Purpose – Class-based storage has been studied extensively and proved to be an efficient storage policy. However, few literature addressed how to cluster stuck items for class-based storage. The purpose of this paper is to develop a constrained clustering method integrated with principal component analysis (PCA) to meet the need of clustering stored items with the consideration of practical storage constraints. Design/methodology/approach – In order to consider item characteristic and the associated storage restrictions, the must-link and cannot-link constraints were constructed to meet the storage requirement. The cube-per-order index (COI) which has been used for location assignment in class-based warehouse was analyzed by PCA. The proposed constrained clustering method utilizes the principal component loadings as item sub-group features to identify COI distribution of item sub-groups. The clustering results are then used for allocating storage by using the heuristic assignment model based on COI. Findings – The clustering result showed that the proposed method was able to provide better compactness among item clusters. The simulated result also shows the new location assignment by the proposed method was able to improve the retrieval efficiency by 33 percent. Practical implications – While number of items in warehouse is tremendously large, the human intervention on revealing storage constraints is going to be impossible. The developed method can be easily fit in to solve the problem no matter what the size of the data is. Originality/value – The case study demonstrated an example of practical location assignment problem with constraints. This paper also sheds a light on developing a data clustering method which can be directly applied on solving the practical data analysis issues.

Constrained Multi-View Video Face Clustering.

In this paper, we focus on face clustering in videos. To promote the performance of video clustering by multiple intrinsic cues, i.e., pairwise constraints and multiple views, we propose a constrained multi-view video face clustering method under a unified graph-based model. First, unlike most existing video face clustering methods which only employ these constraints in the clustering step, we strengthen the pairwise constraints through the whole video face clustering framework, both in sparse subspace representation and spectral clustering. In the constrained sparse subspace representation, the sparse representation is forced to explore unknown relationships. In the constrained spectral clustering, the constraints are used to guide for learning more reasonable new representations. Second, our method considers both the video face pairwise constraints as well as the multi-view consistence simultaneously. In particular, the graph regularization enforces the pairwise constraints to be respected and the co-regularization penalizes the disagreement among different graphs of multiple views. Experiments on three real-world video benchmark data sets demonstrate the significant improvements of our method over the state-of-the-art methods.

A spatial shape constrained clustering method for mammographic mass segmentation.

A novel clustering method is proposed for mammographic mass segmentation on extracted regions of interest (ROIs) by using deterministic annealing incorporating circular shape function (DACF). The objective function reported in this study uses both intensity and spatial shape information, and the dominant dissimilarity measure is controlled by two weighting parameters. As a result, pixels having similar intensity information but located in different regions can be differentiated. Experimental results shows that, by using DACF, the mass segmentation results in digitized mammograms are improved with optimal mass boundaries, less number of noisy patches, and computational efficiency. An average probability of segmentation error of 7.18% for well-defined masses (or 8.06% for ill-defined masses) was obtained by using DACF on MiniMIAS database, with 5.86% (or 5.55%) and 6.14% (or 5.27%) improvements as compared to the standard DA and fuzzy c-means methods.

Hierarchical image representation using 3D camera geometry for content-based image retrieval

In this paper we present a hierarchical image representation methodology by clustering images with 3D camera geometry in order to efficiently retrieve the images according to user׳s viewpoint.The framework of our proposed technique is composed of two steps: first the visual correlation analysis between images in a large database is determined by the estimated 3D camera geometry and second images are classified using a constrained agglomerative hierarchical image clustering method to retrieve the images the users search.The constrained agglomerative hierarchical image clustering method provides balanced hierarchical layers, independent of the number of images within the cluster. It also provides a convenient way to browsing, navigating, and categorizing of the images with various viewpoints, illumination, and partial occlusion.

Improved Color Barycenter Model and Its Separation for Road Sign Detection

This paper proposes an improved color barycenter model (CBM) and its separation for automatic road sign (RS) detection. The previous version of CBM can find out the colors of RS, but the accuracy is not high enough for separating the magenta and blue regions and the influence of number with the same color are not considered. In this paper, the improved CBM expands the barycenter distribution to cylindrical coordinate system (CCS) and takes the number of colors at each position into account for clustering. Under this distribution, the color information can be represented more clearly for analyzing. Then aim to the characteristic of barycenter distribution in CBM (CBM-BD), a constrained clustering method is presented to cluster the CBM-BD in CCS. Although the proposed clustering method looks like conventional K-means in some part, it can solve some limitations of K-means in our research. The experimental results show that the proposed method is able to detect RS with high robustness.

Active co-analysis of a set of shapes

Unsupervised co-analysis of a set of shapes is a difficult problem since the geometry of the shapes alone cannot always fully describe the semantics of the shape parts. In this paper, we propose a semi-supervised learning method where the user actively assists in the co-analysis by iteratively providing inputs that progressively constrain the system. We introduce a novel constrained clustering method based on a spring system which embeds elements to better respect their inter-distances in feature space together with the user-given set of constraints. We also present an active learning method that suggests to the user where his input is likely to be the most effective in refining the results. We show that each single pair of constraints affects many relations across the set. Thus, the method requires only a sparse set of constraints to quickly converge toward a consistent and error-free semantic labeling of the set.

Group RFM analysis as a novel framework to discover better customer consumption behavior

The RFM model provides an effective measure for customers’ consumption behavior analysis, where three variables, namely, consumption interval, frequency, and money amount are used to quantify a customer’s loyalty and contribution. Based on the RFM value, customers can be clustered into different groups and the group information is very useful in market decision making. However, most previous works completely left out important characteristics of purchased products, such as their prices and lifetimes, and apply the RFM measure on all of a customer’s purchased products. This renders the calculation of the RFM value unreasonable or insignificant for customer analysis. In this paper, we propose a new framework called GRFM (for group RFM) analysis to alleviate the problem. The new measure method takes into account the characteristics of the purchased items so that the calculated the RFM value for the customers are strongly related to their purchased items and can correctly reflect their actual consumption behavior. Moreover, GRFM employs a constrained clustering method PICC (for Purchased Items-Constrained Clustering) that could base on a cleverly designed purchase pattern table to adjust original purchase records to satisfy various clustering constraints as well as to decrease re-clustering time. The GRFM allows a customer to belong to different clusters, and thus to be associated with different loyalties and contributions with respect to different characteristics of purchased items. Finally, the clustering result of PICC contains extra information about the distribution status inside each cluster that could help the manager to decide when is most proper to launch a specific sales promotion campaign. Our experiments have confirmed the above observations and suggest that GRFM can play an important role in building a personalized purchasing management system and an inventory management system.

Automatic Urban Water-Body Detection and Segmentation From Sparse ALSM Data via Spatially Constrained Model-Driven Clustering

Identifying hydrological features is important for urban planning and disaster assessment. Data spatial resolution poses challenges in automatic processing. In this letter, we present a novel spatially constrained model-driven clustering method that automatically detects and delineates water bodies in an urban area using airborne laser swath mapping (ALSM) data and imagery. Our method analyzes the modality of the sparseness histogram to decide the existence of water body, followed by clustering. Using the sparseness, clusters are decided by selecting candidate sites. In the iteration of clustering process, new sites are recruited within a close spatial vicinity of the boundary sites. Experiments were conducted using data sets from the city of New Orleans. Our method demonstrated superior robustness regardless of the density of ALSM sample and data discrepancy and very competitive accuracy in comparison with manual tracing, with an overall accuracy above 98%.

Image Annotation Within the Context of Personal Photo Collections Using Hierarchical Event and Scene Models

Most image annotation systems consider a single photo at a time and label photos individually. In this work, we focus on collections of personal photos and exploit the contextual information naturally implied by the associated GPS and time metadata. First, we employ a constrained clustering method to partition a photo collection into event-based subcollections, considering that the GPS records may be partly missing (a practical issue). We then use conditional random field (CRF) models to exploit the correlation between photos based on 1) time-location constraints and 2) the relationship between collection-level annotation (i.e., events) and image-level annotation (i.e., scenes). With the introduction of such a multilevel annotation hierarchy, our system addresses the problem of annotating consumer photo collections that requires a more hierarchical description of the customers' activities than do the simpler image annotation tasks. The efficacy of the proposed system is validated by extensive evaluation using a sizable geotagged personal photo collection database, which consists of over 100 photo collections and is manually labeled for 12 events and 12 scenes to create ground truth.