Bottom-up Clustering Research Articles

Elucidation of Bottom-Up Growth of CaCO3 Involving Prenucleation Clusters from Structure Predictions and Decomposition of Globally Optimized (CaCO3)n Nanoclusters.

Low-energy minima structures for (CaCO3)n, n ≤ 28, are predicted using bottom-up genetic algorithms in conjunction with density functional theory electronic structure calculations, in comparison with the frozen and relaxed top-down clusters generated by cuts from the calcite, vaterite, and aragonite crystal structures. Similarities in structural motifs for the bottom-up and relaxed top-down are revealed using a fragment recognition technique. Fragment energy decomposition analysis shows that the bottom-up and relaxed top-down clusters belong to two classes of amorphous clusters with distinct intracluster energy distributions, despite their structural similarity. The bottom-up clusters with >20 formula units are surface stabilized with negative surface energy densities. In contrast, the top-down clusters are interior stabilized with positive surface energy densities. We prove that the sign of the surface energy density determines whether the nucleation reaction energy as a function of nuclear size has a maximum or a minimum. The surface-stabilized bottom-up clusters are proposed to be a type of prenucleation cluster at the minimum of the nucleation reaction energy. A mechanism for mineralization of CaCO3 involving prenucleation clusters and nonclassical growth pathway is proposed on the basis of our theoretical findings, which is consistent with previous titration experiments.

Clustering of Bacterial Growth Dynamics in Response to Growth Media by Dynamic Time Warping.

Bacterial growth curves, representing population dynamics, are still poorly understood. The growth curves are commonly analyzed by model-based theoretical fitting, which is limited to typical S-shape fittings and does not elucidate the dynamics in their entirety. Thus, whether a certain growth condition results in any particular pattern of growth curve remains unclear. To address this question, up-to-date data mining techniques were applied to bacterial growth analysis for the first time. Dynamic time warping (DTW) and derivative DTW (DDTW) were used to compare the similarity among 1015 growth curves of 28 Escherichia coli strains growing in three different media. In the similarity evaluation, agglomerative hierarchical clustering, assessed with four statistic benchmarks, successfully categorized the growth curves into three clusters, roughly corresponding to the three media. Furthermore, a simple benchmark was newly proposed, providing a highly improved accuracy (~99%) in clustering the growth curves corresponding to the growth media. The biologically reasonable categorization of growth curves suggested that DTW and DDTW are applicable for bacterial growth analysis. The bottom-up clustering results indicate that the growth media determine some specific patterns of population dynamics, regardless of genomic variation, and thus have a higher priority of shaping the growth curves than the genomes do.

MODELING OF THE ORGANIZATION OF INTERACTION OF PARTICIPANTS OF THE MANAGEMENT PROCESS OF CLUSTER COOPERATION DEVELOPMENT

Introduction. Globalization has led to a change in the role of cluster cooperation in economic development. The impact of globalization on regional development of cluster cooperation primarily manifests itself in the fact that it gives regional clusters a new status of subjects of the world economic system and world competitive processes. In the context of globalization, there is a deepening gap in the levels of economic development of regional clusters and an increase in the social vulnerability of the population.Research hypothesis: the concept of regional cluster cooperation covers both a variety of qualitative factors and conditions (informal knowledge networks, trust, image), and a quantitative assessment of attributes and processes (for example, interfirm trade, patenting rates, labor supply, cost of resources used). Regional cluster cooperation consists of the interaction of individual subjects of the region, the social, economic, institutional and social attributes of the region itself. This creates the effect of a combination of factors, increases the competitiveness of the region, creates and (or) uses innovations.The purposeof the study is to substantiate a system of criteria for classifying the forms of spatially localized systems of cluster cooperation, to identify the comparative characteristics of cluster policy models.The research methodology is based on the fundamental principles of a number of scientific fields – the theory of national and regional competitiveness, cluster economic theory, the concept of regional archetypes, theories of spatial proximity, the general theory of systems.Results. The features of the clustering of the Ukrainian economy are identified, the main ones are identified: the formation of bottom-up clusters, discrepancies in the interpretation of the concept of cluster policy, the lack of interest of the cluster members in the interaction.Findings. Analysis of cluster policy in the regions of the EU and Ukraine allowed us to distinguish several types of cluster policy depending on various classification criteria. Depending on the stage of cluster development, the features of cluster policy are highlighted and an effective system of cluster cooperation for the Dnipropetrovsk region is proposed.

A two-stage method for text line detection in historical documents

This work presents a two-stage text line detection method for historical documents. Each detected text line is represented by its baseline. In a first stage, a deep neural network called ARU-Net labels pixels to belong to one of the three classes: baseline, separator or other. The separator class marks beginning and end of each text line. The ARU-Net is trainable from scratch with manageably few manually annotated example images (less than 50). This is achieved by utilizing data augmentation strategies. The network predictions are used as input for the second stage which performs a bottom-up clustering to build baselines. The developed method is capable of handling complex layouts as well as curved and arbitrarily oriented text lines. It substantially outperforms current state-of-the-art approaches. For example, for the complex track of the cBAD: ICDAR2017 Competition on Baseline Detection the F-value is increased from 0.859 to 0.922. The framework to train and run the ARU-Net is open source.

A Bottom-Up Clustering Approach to Unsupervised Person Re-Identification

Most person re-identification (re-ID) approaches are based on supervised learning, which requires intensive manual annotation for training data. However, it is not only resourceintensive to acquire identity annotation but also impractical to label the large-scale real-world data. To relieve this problem, we propose a bottom-up clustering (BUC) approach to jointly optimize a convolutional neural network (CNN) and the relationship among the individual samples. Our algorithm considers two fundamental facts in the re-ID task, i.e., diversity across different identities and similarity within the same identity. Specifically, our algorithm starts with regarding individual sample as a different identity, which maximizes the diversity over each identity. Then it gradually groups similar samples into one identity, which increases the similarity within each identity. We utilizes a diversity regularization term in the bottom-up clustering procedure to balance the data volume of each cluster. Finally, the model achieves an effective trade-off between the diversity and similarity. We conduct extensive experiments on the large-scale image and video re-ID datasets, including Market-1501, DukeMTMCreID, MARS and DukeMTMC-VideoReID. The experimental results demonstrate that our algorithm is not only superior to state-of-the-art unsupervised re-ID approaches, but also performs favorably than competing transfer learning and semi-supervised learning methods.

Synergistic Topology Generation and Route Synthesis for On-Chip Performance-Critical Signal Groups

As very large scale integration technology scales to deep submicron, design for interconnections becomes increasingly challenging. The traditional bus routing follows a sequential bit-by-bit order, and few works explicitly target interbit regularity for signal groups via multilayer topology selection. To overcome these limitations, we present $\mathsf {Streak}$ , an efficient framework that combines topology generation and wire synthesis with a global view of optimization and constrained metal layer track resource allocation. In the framework, an identification stage decomposes binding groups into a set of representative objects; with the generated backbones, equivalent topologies are accompanied by the bits in every object; then a formulation guides the routing considering wire congestion and design regularity. Furthermore, a bottom-up clustering methodology based on layer prediction targets to enhance the routability; a post-refinement stage is developed to match the source-to-sink distance deviation among bits in one group. Experimental results using industrial benchmarks demonstrate the effectiveness of the proposed technique.

The use of long-term features for GMM- and i-vector-based speaker diarization systems

Several factors contribute to the performance of speaker diarization systems. For instance, the appropriate selection of speech features is one of the key aspects that affect speaker diarization systems. The other factors include the techniques employed to perform both segmentation and clustering. While the static mel frequency cepstral coefficients are the most widely used features in speech-related tasks including speaker diarization, several studies have shown the benefits of augmenting regular speech features with the static ones.In this work, we have proposed and assessed the use of voice-quality features (i.e., jitter, shimmer, and Glottal-to-Noise Excitation ratio) within the framework of speaker diarization. These acoustic attributes are employed together with the state-of-the-art short-term cepstral and long-term prosodic features. Additionally, the use of delta dynamic features is also explored separately both for segmentation and bottom-up clustering sub-tasks. The combination of the different feature sets is carried out at several levels. At the feature level, the long-term speech features are stacked in the same feature vector. At the score level, the short- and long-term speech features are independently modeled and fused at the score likelihood level.Various feature combinations have been applied both for Gaussian mixture modeling and i-vector-based speaker diarization systems. The experiments have been carried out on Augmented Multi-party Interaction meeting corpus. The best result, in terms of diarization error rate, is reported by using i-vector-based cosine-distance clustering together with a signal parameterization consisting of a combination of static cepstral coefficients, delta, voice-quality, and prosodic features. The best result shows about 24% relative diarization error rate improvement compared to the baseline system which is based on Gaussian mixture modeling and short-term static cepstral coefficients.

Automatic semantic maps generation from lexical annotations

The generation of semantic environment representations is still an open problem in robotics. Most of the current proposals are based on metric representations, and incorporate semantic information in a supervised fashion. The purpose of the robot is key in the generation of these representations, which has traditionally reduced the inter-usability of the maps created for different applications. We propose the use of information provided by lexical annotations to generate general-purpose semantic maps from RGB-D images. We exploit the availability of deep learning models suitable for describing any input image by means of lexical labels. Lexical annotations are more appropriate for computing the semantic similarity between images than the state-of-the-art visual descriptors. From these annotations, we perform a bottom-up clustering approach that associates each image with a different category. The use of RGB-D images allows the robot pose associated with each acquisition to be obtained, thus complementing the semantic with the metric information.

Segregation of the human basal forebrain using resting state functional MRI

The basal forebrain (BF) is poised to play an important neuromodulatory role in brain regions important to cognition due to its broad projections and complex neurochemistry. While significant in vivo work has been done to elaborate BF function in nonhuman rodents and primates, comparatively limited work has examined the in vivo function of the human BF. In the current study we used multi-echo resting state functional magnetic resonance imaging (rs-fMRI) from 100 young adults (18–34 years) to assess the potential segregation of human BF nuclei as well as their associated projections. Multi-echo processing provided significant gains in SNR throughout the brain as compared to traditional single-echo processing, with some of the largest increases observed in the BF. Bottom-up clustering of voxel-wise BF functional connectivity maps yielded adjacent functional clusters within the BF that closely aligned with the distinct, hypothesized nuclei important to cognition: the nucleus basalis of Meynert (NBM) and the medial septum/diagonal band of Broca (MS/DB). Examining their separate functional connections, the NBM and MS/DB revealed distinct projection patterns, suggesting a conservation of nuclei-specific functional connectivity with homologous regions known to be anatomically innervated by the BF. Specifically, the NBM demonstrated coupling with a widespread cortical network as well as the amygdala, whereas the MS/DB revealed coupling with a more circumscribed network, including the orbitofrontal cortex and hippocampal complex. Collectively, these in vivo rs-fMRI data demonstrate that the human BF nuclei support distinct aspects of resting-state functional networks, suggesting that the human BF may be a neuromodulatory hub important for orchestrating network dynamics.

(a,k)-Anonymous Scheme for Privacy-Preserving Data Collection in IoT-based Healthcare Services Systems.

The widely use of IoT technologies in healthcare services has pushed forward medical intelligence level of services. However, it also brings potential privacy threat to the data collection. In healthcare services system, health and medical data that contains privacy information are often transmitted among networks, and such privacy information should be protected. Therefore, there is a need for privacy-preserving data collection (PPDC) scheme to protect clients (patients) data. We adopt (a,k)-anonymity model as privacy pretection scheme for data collection, and propose a novel anonymity-based PPDC method for healthcare services in this paper. The threat model is analyzed in the client-server-to-user (CS2U) model. On client-side, we utilize (a,k)-anonymity notion to generate anonymous tuples which can resist possible attack, and adopt a bottom-up clustering method to create clusters that satisfy a base privacy level of (a1,k1)-anonymity. On server-side, we reduce the communication cost through generalization technology, and compress (a1,k1)-anonymous data through an UPGMA-based cluster combination method to make the data meet the deeper level of privacy (a2,k2)-anonymity (a1 ≥ a2, k2≥ k1). Theoretical analysis and experimental results prove that our scheme is effective in privacy-preserving and data quality.

Submodular Trajectories for Better Motion Segmentation in Videos.

We propose a new trajectory clustering method using submodular optimization for better motion segmentation in videos. A small number of representative trajectories are first selected by submodular maximization automatically. Then all the initial trajectories can be segmented into fragments with the representative trajectories as centers of fragments. At last, fragments are merged into clusters by a two-stage bottom-up clustering method, and each cluster shows the motion of one moving object. The submodular energy function integrates the quality of all trajectories and their correlations. As a result, thousands of initial trajectories are replaced by only dozens of representative trajectories, which will reduce the negative influence of inaccurate initial trajectories on motion segmentation. The representative trajectories will have larger weights while extracting color or texture information of each moving entity at the step of motion segmentation. Experimental results demonstrate that our method can divide trajectories into more accurate clusters. The final motion segmentation results also illustrate that our method outperforms state-of-the-art motion segmentation methods based on trajectory clustering.

Adopting the Bottom-up Approach and Cluster Analysis on North American and European Male Serial Killers: A Follow-up Study

Adopting the Bottom-up Approach and Cluster Analysis on North American and European Male Serial Killers: A Follow-up Study

Analysis of Massive Scholarly Keywords using Inverted-Index based Bottom-up Clustering

Analysis of Massive Scholarly Keywords using Inverted-Index based Bottom-up Clustering

Active Learning Based Constrained Clustering For Speaker Diarization

Most speaker diarization research has focused on unsupervised scenarios, where no human supervision is available. However, in many real-world applications, a certain amount of human input could be expected, especially when minimal human supervision brings significant performance improvement. In this study, we propose an active learning based bottom-up speaker clustering algorithm to effectively improve speaker diarization performance with limited human input. Specifically, the proposed active learning based speaker clustering has two different stages: explore and constrained clustering. The explore stage is to quickly discover at least one sample for each speaker for boosting speaker clustering process with reliable initial speaker clusters. After discovering all, or a majority, of the involved speakers during explore stage, the constrained clustering is performed. Constrained clustering is similar to traditional bottom-up clustering process with an important difference that the clusters created during explore stage are restricted from merging with each other. Constrained clustering continues until only the clusters generated from the explore stage are left. Since the objective of active learning based speaker clustering algorithm is to provide good initial speaker models, performance saturates as soon as sufficient examples are ensured for each cluster. To further improve diarization performance with increasing human input, we propose a second method which actively select speech segments that account for the largest expected speaker error from existing cluster assignments for human evaluation and reassignment. The algorithms are evaluated on our recently created Apollo Mission Control Center dataset as well as augmented multiparty interaction meeting corpus. The results indicate that the proposed active learning algorithms are able to reduce diarization error rate significantly with a relatively small amount of human supervision.

“Anti-Bayesian” flat and hierarchical clustering using symmetric quantiloids

A Pattern Recognition (PR) system that does not involve labelled samples requires the clustering of the samples into their respective classes before the training and testing can be achieved. All of the reported clustering algorithms (except the one reported in Hammer et al., 2015) operate on Bayesian principles, which is understandable because these principles constitute the basis of optimal PR. Recently, Oommen and his co-authors have proposed a novel, counter-intuitive and pioneering PR scheme that is radically opposed to the Bayesian principle. The rational for this paradigm, referred to as the “Anti-Bayesian” (AB) paradigm, involves classification based on the non-central quantiles of the distributions. This paper, extends the results of Hammer et al. (2015) in many directions. Firstly, we generalize our previous AB clustering Hammer et al. (2015) to handle arbitrary d-dimensional spaces using so-called “quantiloids”. Secondly, we extend the AB paradigm to consider how the clustering can be achieved in hierarchical ways, where we analyze both the Top-Down and the Bottom-Up clustering options. Extensive experimentation, on artificial and on challenging real-life data, demonstrates that our clustering achieves results competitive to the state-of-the-art flat, Top-Down and Bottom-Up clustering approaches, demonstrating the power of the AB paradigm.

Scalable algorithms for unsupervised clustering of acoustic data for speech recognition

In this paper an unsupervised clustering algorithm is developed for acoustic data in the context of speech recognition tasks. One of the key features of the algorithm is scalability to large data sets. Specifically, given the unlabeled training and test sets, the class-labels of the utterances are obtained in an automatic manner. The extracted labels may correspond to the speakers in the speech corpus if the data is relatively clean. The proposed scheme is attractive from an industrial perspective as it alleviates the need to store the speaker-labels manually, saving considerable amount of human efforts and expenses. The core of the algorithm comprises a three-stage architecture that processes the input data one after the other, while each stage is designed to perform a well-defined and specific task. In more detail, the first-pass involves a bottom-up clustering mechanism, the second-pass comprises a cluster splitting operation and the third-pass consists of a cluster refining process. Each of the stages allows for data parallelization using multiple CPUs that leads to faster computation. Two alternative forms of the algorithm are presented – the first considers Gaussian distributions and the other i-Vectors – to facilitate the clustering. Although the algorithm may find applications in various realms of speech recognition, in this paper, the effectiveness of the schemes are evaluated by means of speaker adaptive training (SAT) and speaker-aware training of DNN-HMM acoustic models. In particular, experiments are conducted on the Switchboard task to extract the speaker-labels for the utterances in the training and test sets. It is shown that the SAT DNN-HMM trained using the Gaussian based scheme yields a 7.2% relative improvement in the ASR accuracy over the speaker independent DNN-HMM, whereas the i-Vector approach provides an additional improvement, amounting to a 10.8% relative gain overall. The standard SAT DNN-HMM developed using the ground-truth speaker-labels is found to be only 2.7% relative better than the proposed scheme. Similar observation is made as with speaker-aware training. The analysis of computational complexity, conducted stage by stage, demonstrates the scalability of the proposed algorithms.

Text-Line Detection in Camera-Captured Document Images Using the State Estimation of Connected Components.

Camera-based text processing has attracted considerable attention and numerous methods have been proposed. However, most of these methods have focused on the scene text detection problem and relatively little work has been performed on camera-captured document images. In this paper, we present a text-line detection algorithm for camera-captured document images, which is an essential step toward document understanding. In particular, our method is developed by incorporating state estimation (an extension of scale selection) into a connected component (CC)-based framework. To be precise, we extract CCs with the maximally stable extremal region algorithm and estimate the scales and orientations of CCs from their projection profiles. Since this state estimation facilitates a merging process (bottom-up clustering) and provides a stopping criterion, our method is able to handle arbitrarily oriented text-lines and works robustly for a range of scales. Finally, a text-line/non-text-line classifier is trained and non-text candidates (e.g., background clutters) are filtered out with the classifier. Experimental results show that the proposed method outperforms conventional methods on a standard dataset and works well for a new challenging dataset.

IR based Task-Model Learning: Automating the hierarchical structuring of tasks

Task-models concretize general requests to support users in real-world scenarios. In this paper, we present an IR based algorithm (IRTML) to automate the construction of hierarchically structured task-models. In contrast to other approaches, our algorithm is capable of assigning general tasks close r to the top and specific tasks closer to the bottom. Connections between tasks are established by extending Turney’s PMI-IR measure. To evaluate our algorithm, we manually created a ground truth in the health-care domain consisting of 14 domains. We compared the IRTML algorithm to three state-of-the-art algorithms to generate hierarchical structures, i.e. BiSection K-means, Formal Concept Analysis and Bottom-Up Clustering. Our results show that IRTML achieves a 25.9% taxonomic overlap with the ground truth, a 32.0% improvement over the compared algorithms.

Mechanisms of knowledge flows in bottom-up and top-down cluster initiatives

Knowledge flows are widely believed to be a phenomenon of clusters, and inducing them is one of the chief objectives in establishing and promoting cluster initiatives (CI). However, not many studies discuss how these flows and their effects may differ depending on the mode of CI creation and on the role of public authorities in this process. The main aim of this article is to compare mechanisms of knowledge flows in bottom-up and top-down cluster initiatives. The results of an empirical research involving two case studies in western Poland, obtained through the use of Social Network Analysis (SNA), allowed stating that in bottom-up cluster initiatives firms which were innovation leaders played a prime role in disseminating technological and business knowledge, while in the top-down initiatives the most important were representatives of universities and research centres as well as formal coordinators of cooperation. Policy implications stemming from these results were identified.

Using coevolution to improve protein subfamily classification

Background The common approach for protein subfamily classification relies on grouping protein sequences according to their degree of similarity. However, there is no single sequence similarity threshold for accurately grouping sequences into isofunctional groups. Current subfamily classification methods use bottom-up clustering to construct a cluster hierarchy, then cut the hierarchy at the most appropriate locations to obtain a single partitioning. These methods usually integrate data such as protein sequence similarity, residue conservation within groups and HMM profiles. Despite this straightforward approach, results usually predict a great number of subfamilies with few members and limited biological meaning. The goal of this study is to identify subsets of functionally related sequences within a given superfamily. Since all proteins within a superfamily share a common ancestor, we hypothesize that functional diversity within superfamilies has arisen through a series of concerted changes that must have left an identifiable coevolutionary signal.