Hierarchical Class Research Articles

A Maximum Entropy Framework for Semisupervised and Active Learning With Unknown and Label-Scarce Classes.

We investigate semisupervised learning (SL) and pool-based active learning (AL) of a classifier for domains with label-scarce (LS) and unknown categories, i.e., defined categories for which there are initially no labeled examples. This scenario manifests, e.g., when a category is rare, or expensive to label. There are several learning issues when there are unknown categories: 1) it is a priori unknown which subset of (possibly many) measured features are needed to discriminate unknown from common classes and 2) label scarcity suggests that overtraining is a concern. Our classifier exploits the inductive bias that an unknown class consists of the subset of the unlabeled pool's samples that are atypical (relative to the common classes) with respect to certain key (albeit a priori unknown) features and feature interactions. Accordingly, we treat negative log- p -values on raw features as nonnegatively weighted derived feature inputs to our class posterior, with zero weights identifying irrelevant features. Through a hierarchical class posterior, our model accommodates multiple common classes, multiple LS classes, and unknown classes. For learning, we propose a novel semisupervised objective customized for the LS/unknown category scenarios. While several works minimize class decision uncertainty on unlabeled samples, we instead preserve this uncertainty [maximum entropy (maxEnt)] to avoid overtraining. Our experiments on a variety of UCI Machine learning (ML) domains show: 1) the use of p -value features coupled with weight constraints leads to sparse solutions and gives significant improvement over the use of raw features and 2) for LS SL and AL, unlabeled samples are helpful, and should be used to preserve decision uncertainty (maxEnt), rather than to minimize it, especially during the early stages of AL. Our AL system, leveraging a novel sample-selection scheme, discovers unknown classes and discriminates LS classes from common ones, with sparing use of oracle labeling.

Latent class analysis of bladder urothelial carcinoma to reveal sub-classes defined by alterations to chromatin and signal transduction networks.

469 Background: 70,000 new cases of bladder urothelial carcinoma are diagnosed in the United States each year. Previous studies of this disease suggest two divergent pathways of carcinogenesis, with ~70% of cases being driven by alterations in signal transduction pathways (FGFR3, HRAS) and the remainder featuring alterations in cell cycle genes (TP53, RB1). Other studies focus on the high prevalence of alterations in chromatin remodeling genes (ARID1A, CREBBP, EP300) in this tumor type. The relationship between chromatin modifier alterations and the two bladder TCC sub-classes is not yet understood. Methods: Comprehensive genomic profiling (CGP) on 638 clinically confirmed cases of advanced bladder urothelial carcinoma was performed using the FoundationOne platform. The coding exons of 287 cancer-related and 47 introns of 19 genes frequently rearranged in cancer were sequenced and analyzed for base substitutions, insertions, deletions, copy number alterations and select gene fusions. The resultant alteration profiles were analyzed for their information content with an entropy-based approach and the profiles were subjected to hierarchical latent class analysis (LCA). Results: CGP of 488 male and 150 female UCs of median age 66 (min: 29, max: 87) resulted in 3241 detected alterations across 302 unique genes, for a mean of ~5.1 known or likely somatic alterations per sample. The most commonly altered genes were TP53, CDKN2A/B, ARID1A, KDM6A, and MLL2. Information was concentrated in the 33 most frequently altered genes. Multiple high level clusters were defined by combinatorial patterns of alterations in TP53, CDKN2A/B, RB1, and FGFR3. Sub-classes were defined by mutually exclusive alterations in the chromatin modifiers KDM6A, MLL2, and ARID1A. Conclusions: We present a global picture of UC sub-classes that encompasses previously identified local relationships including the mutual exclusivities between FGFR3 and TP53, RB1 and CDKN2A/B, and KDM6A and MLL2. In the context of clinically relevant genomic alterations that auger response to targeted therapies, identification of sub-classes of UC may further delineate response to therapy.

Latent Condition Qualitative Comparative Analysis: A Mixed Method for Qualitative Analysis

In recent years a great deal of time and effort has been expended toward the development of qualitative methods which retain the advantages of case study and small-n methods, but which ameliorate the various problems (lack of transparency, excessive subjectivity, etc.) Qualitative Comparative Analysis (QCA) is one of the more commonly used examples. QCA allows qualitative researchers to evaluate whether specific factors (conditions) and/or constellations of conditions are necessary, sufficient, or both for a given outcome. While QCA represents an important advance in political science methodology, its primary contributions are confined to how conclusions are drawn from data. Even with a rigorous application of QCA methodology, considerable subjectivity remains in how conditions and outcomes are coded. This is the gap we hope to address with this paper. Taking our inspiration from mixed-methods research design, we develop a new version of QCA, which we call Latent Condition-QCA (LC-QCA). The method relies on the fact that even in studies with a small number of units of analysis, researchers often have a much larger number of units of observation. That is, qualitative coding at higher levels is often based on the aggregation of lower-level units within cases (documents nested with countries, speeches nested within leaders, etc.) In LC-QCA, subjective qualitative coding is conducted at the lowest level of abstraction possible. It is assumed that the probability of observing distinct patterns of these codings is conditional upon the presence or absence of latent conditions. We further assume that conditions at the unit of observation level are predicted by latent conditions at the unit-of-analysis level. We then use hierarchical latent class analysis (HLCA) to assign conditions to units of analysis and observation. These condition assignments are then used to conduct QCA in the normal fashion. The logic used to develop LC-QCA demonstrates a framework for the advancement of social science methods through a true mixture of quantitative and qualitative methods.

Una breve introducción a los modelos de clases jerárquicas (HICLAS)

This article presents a nontechnical introduction to the family of hierarchical classes (HICLAS) models. The original HICLAS model, proposed by De Boeck and Rosenberg in 1988, is suited for the analysis of a binary data matrix; that is, it analyses a two-way table with values equal to 0 or 1. Unlike other classification methods, the output of a HICLAS analysis includes (a) two hierarchical classifications, one for the “objects” (the table's rows) and another for the “attributes” (the columns), and (b) an association of these classifications from which it is possible to predict (or reconstruct) the value in each cell of the table. The advantages of HICLAS models are illustrated with examples in the field of medical education, with an explanation of the different steps in the application of the model, as well as special emphasis on the interpretation of the results.

Zero-Shot Object Recognition System Based on Topic Model

Object recognition systems usually require fully complete manually labeled training data to train the classifier. In this paper, we study the problem of object recognition where the training samples are missing during the classifier learning stage, a task also known as zero-shot learning. We propose a novel zero-shot learning strategy that utilizes the topic model and hierarchical class concept. Our proposed method advanced where cumbersome human annotation stage (i.e. attribute-based classification) is eliminated. We achieve comparable performance with state-of-the-art algorithms in four public datasets: PubFig (67.09%), Cifar-100 (54.85%), Caltech-256 (52.14%), and Animals with Attributes (49.65%) when unseen classes exist in the classification task.

Inefficiency of the Nash equilibrium for selfish machine covering on two hierarchical uniform machines

This paper studies the selfish scheduling game on two hierarchical uniform machines where the jobs are correspondingly classified into two hierarchical classes. The cost of a job is defined as the completion time of the machine to which it is assigned. Each selfish job is interested in minimizing its own cost, while the game seeks to meet the social objective of maximizing the machine cover. We obtain the (strong) price of anarchy and the (strong) price of stability as functions of the ratio between the speeds of the two machines s. We show that all the derived bounds are tight for any value of s, thus completely solving the problem of measuring the inefficiency of the Nash equilibrium on two hierarchical uniform machines.

Hierarchical Classes Analysis (HICLAS): A novel data reduction method to examine associations between biallelic SNPs and perceptual organization phenotypes in schizophrenia

The power of SNP association studies to detect valid relationships with clinical phenotypes in schizophrenia is largely limited by the number of SNPs selected and non-specificity of phenotypes. To address this, we first assessed performance on two visual perceptual organization tasks designed to avoid many generalized deficit confounds, Kanizsa shape perception and contour integration, in a schizophrenia patient sample. Then, to reduce the total number of candidate SNPs analyzed in association with perceptual organization phenotypes, we employed a two-stage strategy: first a priori SNPs from three candidate genes were selected (GAD1, NRG1 and DTNBP1); then a Hierarchical Classes Analysis (HICLAS) was performed to reduce the total number of SNPs, based on statistically related SNP clusters. HICLAS reduced the total number of candidate SNPs for subsequent phenotype association analyses from 6 to 3. MANCOVAs indicated that rs10503929 and rs1978340 were associated with the Kanizsa shape perception filling in metric but not the global shape detection metric. rs10503929 was also associated with altered contour integration performance. SNPs not selected by the HICLAS model were unrelated to perceptual phenotype indices. While the contribution of candidate SNPs to perceptual impairments requires further clarification, this study reports the first application of HICLAS as a hypothesis-independent mathematical method for SNP data reduction. HICLAS may be useful for future larger scale genotype-phenotype association studies.

Comparison of Energy Efficient Data Transmission Approaches for Flat Wireless Sensor Networks

In this paper we have analyzed energy efficient neighbour selection algorithms for routing in wireless sensor networks. Since energy saving or consumption is an important aspect of wireless sensor networks, its precise usage is highly desirable both for the faithful performance of network and to increase the network life time. For this work, we have considered a flat network topology where every node has the same responsibility and capability. We have compared two energy efficient algorithms and analyzed their performances when all sensor nodes in a network have either homogeneous or heterogeneous energy with increase in number of nodes, time rounds and node failures. Wireless sensor networks consist of number of small nodes deployed in an area under supervision. Each node has limited storage, computational and sensing capability and limited energy resource, as nodes are battery operated. Since energy is the main concern in wireless sensor networks (WSN) to maximize the performance and to increase the lifetime of a network, various approaches are implemented to reduce energy consumption in a network. Most of the energy is consumed during idle period and during transmission of data from one node to another i.e. routing. An efficient media access control (MAC) and routing protocol should be designed to save energy. While MAC protocol targets at reduction of energy in scanning and accessing the channel, routing protocol helps to reduce the energy requirement for end-to-end transmission. In WSN, there are number of routing protocols that have been proposed for different network criteria. Based on the network topology WSN protocols have been categorized as - flat network protocols and hierarchical protocols as shown in Figure 1. Protocols that fall under hierarchical class select one head amongst all and form a hierarchy (1). This hierarchy may be a cluster, a chain or a grid. Cluster head or a leader collects the information from all the other nodes in its region and sends it to the sink node or gateway node. Some examples are low energy adaptive clustering hierarchy (LEACH), power efficient gathering in sensor information systems (PEGASIS), virtual grid architecture (VGA), etc. The work presented in this paper considers flat protocols and thus we are not including descriptions of hierarchical protocols and will only be

Mental Health Screening Outcomes Among Justice-Involved Youths Under Community Supervision

The DISC Predictive Scales was administered to 812 New York City youths aged 10–19 placed under community supervision. Approximately half were indicated for possible mental health problems, most frequently mania and posttraumatic stress disorder. Girls were more likely than boys to flag on most disorders. Hierarchical Classes analysis produced five clusters: disruptive behavior, relational distress, marijuana, emotional dysregulation, and specific phobia. Posttraumatic stress disorder and mania were comorbid with all clusters except marijuana. Emotional dysregulation predicted higher, but relational distress predicted lower, risk for rearrest. Marijuana predicted failure to appear in court and receiving a final disposition of placement.

Optimal algorithms for semi-online machine covering on two hierarchical machines

This paper investigates the semi-online machine covering problem on two hierarchical machines where the jobs are correspondingly classified into two hierarchical classes. The objective is to maximize the minimum machine load. We show that if we only know the size of the largest job, no algorithm with a bounded competitive ratio exists. So we consider the case where we know both the size and the class of the largest job. If we know the size of the largest job and that it belongs to the higher class, then an optimal algorithm with a (1+22)-competitive ratio exists. If we know the size of the largest job and that it belongs to the lower class, we design an optimal algorithm with an α-competitive ratio, where α≈2.48119 is the largest root of the equation x3−2x2−2x+2=0. For the case where the total size of all the jobs is known in advance, we show that the competitive ratio of an optimal algorithm is 2.

Mismatches between objective parameters and measured perception assessment in room acoustics: A holistic approach

Psychoacoustic research in the field of concert halls has revealed that many aspects concerning listening perception have yet to be totally understood. On the one hand, the objective room acoustics of performance spaces are reflected in parameters, some standardized and some not, but these are related to a limited number of perceptual attributes of human response. In general, these objective parameters cannot accurately describe the acoustic details due to their inherent simplification. Under these premises, impulse responses (576 receivers) are measured in 16 concert halls, according to standard procedures, and the perception and satisfaction of the occupants of the rooms are evaluated by completing a questionnaire during live concerts. Correlation analyses and multidimensional scaling (MDS) techniques have been applied to spatial and multi-band averaged values of the acoustic parameters studied (18), and the average values of users' responses (1284) to the questionnaire items (26). As a first result, correlations between objective parameters and users' responses show that transversality exists between them. Secondly, hierarchical clustering produces the classification of survey questions in 7 hierarchical classes. On the other hand, a lack of tuning between objective parameters and perceptual responses is observed on applying MDS analysis to the ordination of the venues from a subjective assessment and a subjective-objective assessment. Finally, although the results show the mismatch between objective parameters and subjective responses, a model of subjective global evaluation of the acoustics of the room from data of three orthogonal acoustic parameters is implemented, revealing a reasonably good fit.

UML Models of Design and Knowledge Representation for Technical Production Preparation Needs

Having in mind the need of the computer adding of the technical production preparation functions the elaboration of models of the design and knowledge representation is essential. In order to prepare these models the object paradigm in the meaning of object oriented techniques and the UML language would be applied. Thanks to this solution the coherent representation of the design and manufacturing knowledge in the framework of a new created CAPP is possible to achieve. The main purpose of the presented work is to show and discuss original models of a design and manufacturing knowledge representations. In the proposed solution the product design is represented by means of the hierarchical class structure in which a particular class is responsible for representing a characteristic type of the model feature. The manufacturing knowledge similarly to knowledge representation is also represented by means of the hierarchical class structure. Each class in the manufacturing knowledge model stores a certain amount of manufacturing knowledge. This model allow representing both procedural and declarative knowledge. The declarative character of the model results from storing of manufacturing process planning rules whilst procedural one from recording of manufacturing instrumentation, cutting parameters etc. selection procedures. Having these models done it is possible to undertake a research on connections between a design and manufacturing knowledge representations.

Examination of categorical approach to symptom assessment: cross-validation of foulds’ Delusions-Symptoms-States Inventory with Korean non-patient and patient groups

BackgroundFoulds’ Delusions-Symptoms-State Inventory (DSSI) has been purported to be a reliable, systematic categorical measure to assess the patients with schizophrenia according to the degree of illness. However, further cross-validations using other clinical measures and diverse samples from other cultures have not been advanced recently. We aimed to examine the validity of the DSSI hierarchical class model using both Korean non-patient and patient (schizophrenia and depression) groups.MethodThe hypothesis of inclusive, non-reflexive relationships among the DSSI classes was tested. The power of DSSI to detect presence of symptoms was assessed via cross-validation with other clinical measures, and the differences between the clinical features among the DSSI classes were examined using the Minnesota Multiphasic Personality Inventory (MMPI).ResultsThe high rate of model conformity (91.1%) across the samples and cross-validation with other criterion measures provided further support for the validity of DSSI.ConclusionsDSSI is a reliable self-report measure that can be applied to both patient and non-patients to assess the presence and severity of psychiatric illness. Future studies that include more diverse clinical groups are necessary to lend further support for its utility in clinical practice.

Revealing interdyad differences in naturally occurring staff reactions to challenging behaviour of clients with severe or profound intellectual disabilities by means of Clusterwise Hierarchical Classes Analysis (HICLAS)

Investigating interdyad (i.e. couples of a client and their usual caregiver) differences in naturally occurring patterns of staff reactions to challenging behaviour (e.g. self-injurious, stereotyped and aggressive/destructive behaviour) of clients with severe or profound intellectual disabilities is important to optimise client-staff interactions. Most studies, however, fail to combine a naturalistic setup with a person-level analysis, in that they do not involve a careful inspection of the interdyad differences and similarities. In this study, the recently proposed Clusterwise Hierarchical Classes Analysis (HICLAS) method is adopted and applied to data of in which video fragments (recorded in a naturalistic setting) of a client showing challenging behaviour and the staff reacting to it were analysed. In a Clusterwise HICLAS analysis, the staff-client dyads are grouped into a number of clusters and the prototypical behaviour-reaction patterns that are specific for each cluster (i.e. interdyad differences and similarities) are revealed. Clusterwise HICLAS discloses clear interdyad differences (and similarities) in the prototypical patterns of clients' challenging behaviour and the associated staff reactions, complementing and qualifying the results of earlier studies in which only general patterns were disclosed. The usefulness and clinical relevance of Clusterwise HICLAS is demonstrated. In particular, Clusterwise HICLAS may capture idiosyncratic aspects of staff-client interactions, which may stimulate direct support workers to adopt person-centred support practices that take the specific abilities of the client into account.

Translating land cover/land use classifications to habitat taxonomies for landscape monitoring: a Mediterranean assessment

Periodic monitoring of biodiversity changes at a landscape scale constitutes a key issue for conservation managers. Earth observation (EO) data offer a potential solution, through direct or indirect mapping of species or habitats. Most national and international programs rely on the use of land cover (LC) and/or land use (LU) classification systems. Yet, these are not as clearly relatable to biodiversity in comparison to habitat classifications, and provide less scope for monitoring. While a conversion from LC/LU classification to habitat classification can be of great utility, differences in definitions and criteria have so far limited the establishment of a unified approach for such translation between these two classification systems. Focusing on five Mediterranean NATURA 2000 sites, this paper considers the scope for three of the most commonly used global LC/LU taxonomies—CORINE Land Cover, the Food and Agricultural Organisation (FAO) land cover classification system (LCCS) and the International Geosphere-Biosphere Programme to be translated to habitat taxonomies. Through both quantitative and expert knowledge based qualitative analysis of selected taxonomies, FAO-LCCS turns out to be the best candidate to cope with the complexity of habitat description and provides a framework for EO and in situ data integration for habitat mapping, reducing uncertainties and class overlaps and bridging the gap between LC/LU and habitats domains for landscape monitoring—a major issue for conservation. This study also highlights the need to modify the FAO-LCCS hierarchical class description process to permit the addition of attributes based on class-specific expert knowledge to select multi-temporal (seasonal) EO data and improve classification. An application of LC/LU to habitat mapping is provided for a coastal Natura 2000 site with high classification accuracy as a result.

Algorithm to Match Ontologies on the Semantic Web

It has been recognized that semantic data and knowledge extraction will significantly improve the capability of natural language interfaces to the semantic search engine. Semantic Web technology offers a vast scale of sharing and integration of distributed data sources by combining information easily. This will enable the user to find the information easily and efficiently. In this paper, we will explore some issues of developing algorithms for the Semantic Web. The first one to build the semantic contextual meaning by scanning the text, extract knowledge and automatically infer the meaning of the information from text that contains the search words in any sentence and correlate with hierarchical classes defined in the Ontology as a result of input resources. The second to discover the hierarchical relationships among terms (i.e. discover the semantic relations across hierarchical classifications). The proposed algorithm will be relying on a number of resources including Ontology and WordNet.

Quantifying the Hierarchical Organization of Compensation In Submarine Fans Using Surface Statistics

Abstract Stratigraphy is often interpreted within hierarchical, or scale-dependent, frameworks that subdivide deposits based on distinct jumps in characteristics such as duration of deposition or scale. While the interpretation is logically valid, few studies quantitatively demonstrate that the jumps exist. Rather, recent work has quantitatively shown some characteristics of stratigraphy to be fractal, or scale invariant. Compensational stacking, the tendency for sediment-transport systems to preferentially fill topographic lows, is a concept widely used in stratigraphic interpretation. Here we use the compensation index, a metric that quantifies the strength of compensational stacking in sedimentary deposits, to describe the architecture of stratigraphy exposed in outcrops of submarine-fan strata in the Carboniferous Ross Sandstone representing contrasting architectural styles: (1) predominantly lobe elements and (2) predominantly channel elements. In both datasets, the stratigraphic architecture is classified into hierarchical classes of beds, stories, and elements. Results are the following. First, at both sites we document statistically significant increases in the strength of compensation across larger hierarchical levels supporting the use of hierarchical interpretations of stratigraphy. It is therefore plausible for some characteristics of sedimentary systems to be hierarchical and others to be fractal. Second, we document that lobe elements stack more compensationally than channel elements. We interpret this pattern to document that compensation increases along a longitudinal transect through this distributive submarine fan.

Accessing child developmental services predicts in-home placement of substance- and HIV-affected children

Abstract Intake and termination data from two samples of substance-using and/or HIV-positive mothers of minor children were collected from 17 projects across the U.S. Service utilization data collected from the initial sample was modeled using hierarchical classes (HICLAS) analysis. HICLAS identified three basic classes of service utilization: Services to Meet Basic Needs , Treatment and Support Services , and Pediatric and Developmental Services . After computing propensity score adjustments, we evaluated whether these service classes predicted child placement at termination: living with biological mother without Child Protective Service (CPS) involvement, living with biological mother with CPS involvement, or not living with biological mother. In both the initial and the replication samples, participating mothers who received Services to Meet Basic Needs and/or Pediatric and Developmental Services were more likely to be living with their index children without CPS involvement. Pediatric and Developmental Services was the only significant predictor of this outcome for the subsample of biological mothers who were not living with their index children at program intake. Findings are discussed in terms of the possible benefits derived from the process of receiving services in the Pediatric and Developmental Services category.

Addressing identity crime in crime management information systems: Definitions, classification, and empirics

Identity fraud as a term and concept in its formative stages was often presumed to be identity theft and visa versa. However, identity theft is caused by the identities (or tokens) of individuals or organisations being stolen is an enabling precursor to identity fraud. The boundaries of identity fraud and identity theft are now better defined. The absence of specific identity crime legislation could be a cause of perpetrators not classified as breaching identity crimes but under other specific entrenched law such as benefit fraud, or credit card fraud. This metrics overlap can cause bias in crime management information systems. This study uses a multi-method approach where data was collected in both a quantitative and qualitative manner. These approaches are used as a lens for defining different classes of online identity crimes in a crime management (IS) security context. In doing so, we contribute to a deeper understanding of identity crime by specifically examining its hierarchical classes and definitions; to aid clearer structure in crime management IS. We seek to answer the questions: should current law around identity fraud continue to be reinforced and measures introduced to prevent identity crime; should laws be amended; or should new identity crime laws be constructed? We conclude and recommend a solution incorporating elements of all three.

Ordination and hyperspectral remote sensing approach to classify peatland biotopes along soil moisture and fertility gradients

Spatial information on biotopes would be an asset for boreal peatland inventories. It may be helpful for protection and monitoring changes in species diversity caused by artificial regulation of hydrology, atmospheric deposition, and global change. We took a data-driven approach to explore the potential of airborne hyperspectral data in determining plant associations of treeless northern boreal peatlands in Finland (65°57′N, 24°29′E). The objective was to produce an ecological peatland classification on a class hierarchical level which would be as detailed as possible, and could be accurately mapped with remotely sensed hyperspectal data. The proper hierarchical class level was decided based on statistical approaches including canonical correspondence analysis, unsupervised classification, and multivariate analysis of variance. Subsequently, fuzzy classification with support vector machines was applied to map the classes in the spatial domain from airborne geo- and atmospherically corrected hyperspectal HyMap data. We conducted detailed in situ sampling for surface plant inventory, soil root zone bulk dielectric permittivity (ε), electrical conductivity (σ), and pH. Constrained canonical correspondence ordination revealed that ε, σ (p<0.01), and pH (p<0.001) were significantly correlated with the ordination and were thus important drivers for vegetation composition. Four biotopes were distinguishable by their spectra: Sphagnum fuscum–Emperum nigrum–Rubus chamaemorus–S. angustifolium-bog, Eriophorum vaginatum–S. lindenbergii-sedge fen, eutrophic Carex lasiocarpa–Trichophorum cespitosum-fen, and Comarum palustre–S. majus–Carex rostrata-swamp. The highest overall efficiencies of prediction (overall accuracy 87.8%, kappa 0.781) were calculated using only 10% of the original training data indicating that low numbers of properly selected training sites from class boundaries could be appropriate for the support vector machines. The class specific assessment of accuracy with receiver operating characteristics curves also confirmed moderate-to-good successes of the approach with area under the curve values: 0.946 for bog, 0.951 for sedge fen, and 0.999 for eutrophic fen. Clusters of spectrally separable cohesive communities existed but they gradually changed from one plant association to another. Thus fuzzy mapping from hyperspectral and possible complementary data with well performing support vector machines could be justified in spatially continuous inventory and monitoring of boreal peatlands at regional, landscape or national scales.