Cumulative Classification Research Articles

Dynamic Loss Reweighting Method Based on Cumulative Classification Scores for Long-Tailed Remote Sensing Image Classification

Convolutional neural networks have been widely used in remote sensing classification and achieved quite good results. Most of these methods are based on datasets with relatively balanced samples, but such ideal datasets are rare in applications. Long-tailed datasets are very common in practice, and the number of samples among categories in most datasets is often severely uneven and leads to bad results, especially in the category with a small sample number. To address this problem, a novel remote sensing image classification method based on loss reweighting for long-tailed data is proposed in this paper to improve the classification accuracy of samples from the tail categories. Firstly, abandoning the general weighting approach, the cumulative classification scores are proposed to construct category weights instead of the number of samples from each category. The cumulative classification score can effectively combine the number of samples and the difficulty of classification. Then, the imbalanced information of samples from each category contained in the relationships between the rows and columns of the cumulative classification score matrix is effectively extracted and used to construct the required classification weights for samples from different categories. Finally, the traditional cross-entropy loss function is improved and combined with the category weights generated in the previous step to construct a new loss reweighting mechanism for long-tailed data. Extensive experiments with different balance ratios are conducted on several public datasets, such as HistAerial, SIRI-WHU, NWPU-RESISC45, PatternNet, and AID, to verify the effectiveness of the proposed method. Compared with other similar methods, our method achieved higher classification accuracy and stronger robustness.

Braitenberg Vehicles as Developmental Neurosimulation.

Connecting brain and behavior is a longstanding issue in the areas of behavioral science, artificial intelligence, and neurobiology. As is standard among models of artificial and biological neural networks, an analogue of the fully mature brain is presented as a blank slate. However, this does not consider the realities of biological development and developmental learning. Our purpose is to model the development of an artificial organism that exhibits complex behaviors. We introduce three alternate approaches to demonstrate how developmental embodied agents can be implemented. The resulting developmental Braitenberg vehicles (dBVs) will generate behaviors ranging from stimulus responses to group behavior that resembles collective motion. We will situate this work in the domain of artificial brain networks along with broader themes such as embodied cognition, feedback, and emergence. Our perspective is exemplified by three software instantiations that demonstrate how a BV-genetic algorithm hybrid model, a multisensory Hebbian learning model, and multi-agent approaches can be used to approach BV development. We introduce use cases such as optimized spatial cognition (vehicle-genetic algorithm hybrid model), hinges connecting behavioral and neural models (multisensory Hebbian learning model), and cumulative classification (multi-agent approaches). In conclusion, we consider future applications of the developmental neurosimulation approach.

Intertwining Binary Decision Trees and Probabilistic Neural Networks for Maximising Accuracy and Efficiency in Classification Tasks A Pilot /Proof–of–Concept Demonstration on the Iris Benchmark Classification Dataset

Intertwining binary decision trees (BDTs) and probabilistic neural networks (PNNs) is put forward as a powerful custom–made methodology for simultaneously maximising accuracy and efficiency in classification tasks. The proposed methodology brings together the complementary strengths of (I) the parametric, global, recursive, efficient as well as maximal dichotomisation of the BDT, and (II) the non–parametric, accurate–to–local–detail, multi–class identification of the PNN. The BDT/PNN combination evolves level–by–level, with each level comprising two steps: (step 1) the optimal BDT (hyper–)linear bi–partitioning of the entire dataset for the first level, and of every non–terminal partition thereof for the next levels, is determined, with each resulting partition being assigned to a new node of the next BDT level; (step 2) as many PNNs are created as there are  multi–class partitions resulting from (step 1), implementing the non–parametric creation of the hyperplane that maximises class separability over each corresponding partition of the problem–space. BDT/PNN expansion is applied recursively – and independently – to each non–terminal partition of the dataset, concluding once the cumulative classification accuracy (CCA) of the terminal – BDT and PNN – nodes of the BDT/PNN is maximised. A proof–of–concept demonstration on the iris benchmark classification dataset (IBCD) illustrates the details of the BDT/PNN combination, attesting to its simplicity, effectiveness, and efficiency of operation. Follow–up research shall focus upon the classification of additional benchmark datasets, as well as of “hard” and “Big” real–world problems, for further evaluating and validating the proposed methodology.

Facial Recognition for Drunk People Using Thermal Imaging

Face recognition using thermal imaging has the main advantage of being less affected by lighting conditions compared to images in the visible spectrum. However, there are factors such as the process of human thermoregulation that cause variations in the surface temperature of the face. These variations cause recognition systems to lose effectiveness. In particular, alcohol intake causes changes in the surface temperature of the face. It is of high relevance to identify not only if a person is drunk but also their identity. In this paper, we present a technique for face recognition based on thermal face images of drunk people. For the experiments, the Pontificia Universidad Católica de Valparaíso-Drunk Thermal Face database (PUCV-DTF) was used. The recognition system was carried out by using local binary patterns (LBPs). The LBP features were obtained from the bioheat model from thermal image representation and a fusion of thermal images and a vascular network extracted from the same image. The feature vector for each image is formed by the concatenation of the LBP histogram of the thermogram with an anisotropic filter and the fused image, respectively. The proposed technique has an average percentage of 99.63% in the Rank-10 cumulative classification; this performance is superior compared to using LBP in thermal images that do not use the bioheat model.

Evolutionary extreme learning machine with sparse cost matrix for imbalanced learning

Extreme learning machine is a popular machine learning technique for single hidden layer feed-forward neural network. However, due to the assumption of equal misclassification cost, the conventional extreme learning machine fails to properly learn the characteristics of the data with skewed category distribution. In this paper, to enhance the representation of few-shot cases, we break down that assumption by assigning penalty factors to different classes, and minimizing the cumulative classification cost. To this end, a case-weighting extreme learning machine is developed on a sparse cost matrix with a diagonal form. To be more actionable, we formulate a multi-objective optimization with respect to penalty factors, and optimize this problem using an evolutionary algorithm combined with an error bound model. By doing so, this proposed method is developed into an adaptive cost-sensitive learning, which is guided by the relation between the generalization ability and the case-weighting factors. In a broad experimental study, our method achieves competitive results on benchmark and real-world datasets for software bug reports identification.

АНАЛИЗ СТРУКТУРНЫХ ОСОБЕННОСТЕЙ ЭПОНИМОВ И ТОПОНИМОВ В АНГЛОЯЗЫЧНОМ НАУЧНО-ТЕХНИЧЕСКОМ ДИСКУРСЕ

The author of this article studies the eponymic and toponymic structures of the English scientific-technical discourse (the sphere of nanotechnologies and fiber-optic technology). In the article the features of structural models of eponyms and toponyms are considered, the differentiation of the indicated terms is carried out, the complex classification of eponyms and toponyms according to their structure is conducted, their cumulative classification is proposed proceeding from the quantitative structure of the structural model. The author of the article offers a detailed description of each distinguished structure, conducts a statistical analysis of the considered examples, draws a conclusion about the most frequent toponymic-eponymic model in the scientific-technical discourse.

The practice of applying the provision of Article 297 of the Polish Penal Code

The aim of the publication is to present the results of research on the issue of the application of the provision of Article 297 of Polish Penal Code in practice and an attempt to evaluate them and an attempt to formulate conclusions. The features characterizing the offense under Article 297 § 1 of Polish Penal Code, which reads as follows: who, in order to obtain for himself or someone else, from a bank or organizational unit conducting similar economic activity under the Act or from an authority or institution with public funds - credit, monetary loan, surety, guarantee, letter of credit, subsidy, subsidy, confirmation by the bank of a liability arising from a surety or a guarantee or similar cash payment for a specific business purpose, payment instrument or public contract, submit counterfeit, converted, attesting untruth or unreliable document or unreliable written statement regarding circumstances of significant importance in order to obtain the mentioned financial support, payment instrument or order, it is punishable by imprisonment up from 3 months to 5 years. Discussed are also the indications of committing an act consisting in abandoning the notification in Article 297 § 2 of Polish Penal Code, worded as follows: who, in breach of a binding obligation, does not notify the relevant entity of the situation that may affect the cessation or limitation of the financial support referred to in § 1 or the public order or the possibility of further use of the electronic payment instrument. Next, the methodology of the research was described in detail. The research was of an files nature and concerned final judgments issued by courts of second instance. The time period of researches was from 2012 to 2016. A purposeful selection was also made due to the legal qualification of the deed on the basis of which the conviction was handed down in the first instance, i.e. the offenses under Article 297 of Polish Penal Code. Then the results of the research were presented and an attempt was made to comment on the results obtained. The conclusions from the research indicated, among others, the cases of lack of correct understanding of the provision, confusion of concepts, unnecessary automatic cumulative classification of the deed, narrowing the use of the instruments mentioned in the provision to credits, loans and subsidies, and many more cases.

Late Presentation of SLE and Secondary Anti-phospholipid Syndrome, Following Initial Evans Syndrome

Background: Evans syndrome associated with secondary anti-phospholipid syndrome has not been reported in case studies and we report a case of SLE and secondary anti-phospholipid syndrome diagnosed seven years after the initial diagnosis of Evans syndrome. Case presentation: A 21-years-old female was diagnosed with Evans syndrome since the age of 14 years, and was on treatment for seven years. At the onset of disease her immunological screening for ANA, Ds-DNA were negative. She was treated with immunosuppressants and steroids and the disease was complicated with Cryptococcal meningitis. Seven years later at the age of 21 years she started to have joint arthritis, and the ANA test converted to positive together with a positive Ds-DNA. Later on she had a single episode of unprovoked deep vein thrombosis in a leg (below knee), and became positive for lupus anticoagulant and IgM anticardiolipin antibodies. Her cardiac imaging (Echo-cardiogram) revealed that she had developed moderate pulmonary hypertension. She was treated with physiotherapy, Methotrexate, Hydroxychloroquine, Warfarin and Sildenafil with good control of the disease. Her lupus is in remission with treatment with no anaemia. Conclusion: Haematological manifestations of systemic lupus erythematosus could vary. Hence, Lupus should be kept in mind always during management of complicated hematological conditions as the patient might fulfill cumulative SLE classification criteria over time.

Simultaneous fecal microbial and metabolite profiling enables accurate classification of pediatric irritable bowel syndrome

BackgroundWe previously showed that stool samples of pre-adolescent and adolescent US children diagnosed with diarrhea-predominant IBS (IBS-D) had different compositions of microbiota and metabolites compared to healthy age-matched controls. Here we explored whether observed fecal microbiota and metabolite differences between these two adolescent populations can be used to discriminate between IBS and health.FindingsWe constructed individual microbiota- and metabolite-based sample classification models based on the partial least squares multivariate analysis and then applied a Bayesian approach to integrate individual models into a single classifier. The resulting combined classification achieved 84 % accuracy of correct sample group assignment and 86 % prediction for IBS-D in cross-validation tests. The performance of the cumulative classification model was further validated by the de novo analysis of stool samples from a small independent IBS-D cohort.ConclusionHigh-throughput microbial and metabolite profiling of subject stool samples can be used to facilitate IBS diagnosis.Electronic supplementary materialThe online version of this article (doi:10.1186/s40168-015-0139-9) contains supplementary material, which is available to authorized users.

Robust multi-scale ship tracking via multiple compressed features fusion

In this paper, we address the problem of tracking a single ship in inland waterway closed circuit television (CCTV) video sequences given its location in the first frame and no other prior information. First, based on the compressive sensing theory, we employ two kinds of random measurement matrices to extract two complementary good features to track the target ship. Second, in order to track both location and scale, we construct our random measurement matrices according to spatial and temporal structure constraints in consecutive frames, which can be easily obtained and recorded in an offline manner. Having obtained the low-dimensional features in the compressed domain, we further take the different discriminability strengths of the extracted features into account and perform feature evaluations through their cumulative classification performances. A naive Bayes classifier with online update is employed to determine whether the image patch belongs to the foreground or background and a coarse-to-fine strategy is adopted to speed up the time-consuming detection procedure. Finally, both qualitative and quantitative evaluations on numerous challenging CCTV videos demonstrate that the proposed algorithm outperforms several state-of-the-art methods in terms of accuracy, precision and robustness

The Impact of Multiple Types of Child Maltreatment on Subsequent Risk Behaviors Among Women During the Transition from Adolescence to Young Adulthood

The purpose of this study was to investigate how different types of child maltreatment, independently and collectively, impact a wide range of risk behaviors that fall into three domains: sexual risk behaviors, delinquency, and suicidality. Cumulative classification and Expanded Hierarchical Type (EHT) classification approaches were used to categorize various types of maltreatment. Data were derived from Wave III of the National Longitudinal Study of Adolescent Health (Add Health). Our sample consisted of White, Black, Hispanic, and Asian females ages 18 to 27 (n = 7,576). Experiencing different kinds of maltreatment during childhood led to an extensive range of risk behaviors within the three identified domains. Women experiencing sexual abuse plus other maltreatment types had the poorest outcomes in all three domains. These findings illustrate that it may no longer be appropriate to assume that all types of maltreatment are equivalent in their potential contribution to negative developmental sequelae.

Effects of multiple maltreatment experiences among psychiatrically hospitalized youth

Objective Relying on indicators coded from information collected routinely during intake assessments at a secure inpatient psychiatric facility, this study examined the extent to which different forms of maltreatment accounted for variations in youths’ emotional and behavioral problems. Methods Clinical information was reviewed for a large ( N = 401) and diverse sample (mean age = 13.9 years; 53% male; 54.6% racial/ethnic minority) of youth admitted to a publicly funded psychiatric hospital. Data were drawn from intake narratives, standardized psychopathology rating scales, and psychiatric diagnostic ratings. Results Findings provide some support for a hierarchical classification of multiple maltreatment experiences with sexual abuse identified as the specific form of maltreatment most reliably associated with poor adjustment. Support also was obtained for a cumulative classification approach, as the number of different types of maltreatment experiences was linked positively to elevated psychopathology ratings. Conclusions Even in this high-risk, atypical sample, maltreatment experiences account for variation in levels of psychopathology. These results have implications for classifying multiple maltreatment and enhancing clinical care for atypical youth who have been maltreated. Practice implications Clinicians working in youth psychiatric populations should implement maltreatment-specific psychotherapy approaches for maltreated youth, even as adjunctive treatments in a therapeutic milieu.

Gene Expression Profiling Predicts Outcome in De Novo Acute Myeloid Leukemia (AML) with Normal Karyotype: Results of Children's Oncology Group (COG) Study POG #9421.

We previously reported a 36% event-free survival (EFS) estimate for patients with normal karyotype (NK) on the COG study POG #9421 (n=144). In addition, we hypothesized that gene expression profiling would identify signatures linked to clinical outcome and useful for retrospective risk determination. Bone marrows in a subset of patients with NK (n=58) were analyzed using a 43,760-element spotted arrays containing 41,751 unique genes and expressed sequence tags; arrays were obtained from the Stanford University Microarray Core Facility. Prediction analysis for microarrays (PAM) was used to find genes that identified samples associated-with and unassociated-with events (relapse or death); after analyzing 28,711 genes with PAM we chose a 727-gene cluster that differentiated patients with NK on the basis of clinical outcome (cumulative classification error rate 19%). The analysis was biased for a larger number of genes in order to obtain a more biologically informative gene pathways analysis. Significance analysis of microarrays (SAM) on the PAM output identified 633 genes (false-discovery rate of 0%) that differed significantly between the event-associated and event-unassociated samples. Spearman based hierarchical clustering on these genes yielded 2 clusters with statistically significant different event-free survivals: 65% (n=24) for the event-unassociated curve and 23% (n=34) for the event-associated curve with P=0.01.The patients in these clusters did not differ at diagnosis for WBC (70K vs. 100K/microL with P=0.19) and age (10.1 vs. 9.9 yrs with P=0.83) by unpaired t-test; or for sex (P=0.11) and FLT3-ITD status (P=0.76) by Fisher's exact test. The gene list (GenBank #) and fold-change in gene expression from SAM output were analyzed using Ingenuity Pathway Analysis software (Ingenuity™ Systems, Mountain View, CA). Canonical pathways identified 33 networks associated with event-unassociated outcome using 302 eligible genes that included: underexpressed CDC73, RAD50, SPARC, PTPN12, MXD1, TNF, ABCA1, STAT4, CCNA1, TNF, BCL2A1, JUN, BCL6 and AREG; and overexpressed RUNX3, FKBP9, FKBP8, MAP2K2, CHES1, HOXA11, HRK, CDK6, MGA, MAPK3, ABL1, HDAC7A, SMARCC2, SYK, MXD4, CDC42. Several of these genes have been previously reported to be associated with improved outcome in AML. However, two of these genes (MXD4 and MXD1) previously not identified as related to outcome in AML, but identified in our analysis in two highly interacting networks related to the MYC gene, result in a difference in EFS of 51% vs. 24% (P=0.04), suggesting that a smaller number of genes may be predictive of outcome.Conclusion: Risk assignment for patients with NK may be feasible by analyzing a limited number of genes. We will validate these findings by correlating gene expression results with quantitative real-time PCR. Prospective validation of this strategy in clinical trials may be warranted.

New Methods for the Analysis of Binarized BIOLOG GN Data of Vibrio species: Minimization of Stochastic Complexity and Cumulative Classification

We apply minimization of stochastic complexity and the closely related method of cumulative classification to analyse the extensively studied BIOLOG GN data of Vibrio spp. Minimization of stochastic complexity provides an objective tool of bacterial taxonomy as it produces classifications that are optimal from the point of view of information theory. We compare the outcome of our results with previously published classifications of the same data set. Our results both confirm earlier detected relationships between species and discover new ones.

Bayesian Predictive Identification and Cumulative Classification of Bacteria

In this paper we give a mathematically precise formulation of an old idea in bacterial taxonomy, namely cumulative classification, where the taxonomy is continuously updated and possibly augmented as new strains are identified. Our formulation is based on Bayesian predictive probability distributions. The criterion for founding a new taxon is given a firm theoretical foundation based on prediction and it is given a clear-cut interpretation. We formulate an algorithm for cumulative classification and apply it to a large database of bacteria belonging to the familyEnterobacteriaceae. The resulting taxonomy makes microbiological sense.

An Assessment of Cumulative Classification

We present a method for building systematics when new knowledge is continuously accumulated. The resulting classification is self-correcting and improves itself by sorting new items as they are added to the material and studied. The formulation is based on Bayesian predictive probability distributions. A new item that has not yet been classified is assigned to the class that has maximal posterior probability or is made to form a group of its own. Such a cumulative classification depends on the order in which the items are classified. The introduction of an already classified training set considerably improves the repeatability of the method. As a case study we applied the method to a large data set for the Enterobacteriaceae. The resulting classifications corresponded well to the general structure of the prevailing taxonomy of Enterobacteriaceae.

Online recognition by deviation-expansion model and A ∗ algorithm-based matching

This paper presents an online recognition system for large-alphabet handprinted Chinese characters using a model-based recognition approach with stroke-based features. A deviation-expansion (D-E) model representing the reference pattern is constructed. The model contains hypothetical knowledge of handwriting variations, including stroke-order deviations and strokenumber deviations. For pattern matching a matching tree is constructed by combining the knowledge of the reference pattern and the unknown pattern together. With the tree a similarity measure function is defined to indicate the degree of similarity. Evaluation of the function is obtained using A ∗ algorithm-based matching. Experimental results are based upon testing a set of 54010 handprinted sample characters written in the square style by ten people. The cumulative classification rate of choosing the ten most similar characters is 98%. The results suggest that the hypothetical model is both feasible and reasonable.

On-line recognition by deviation-expansion model and dynamic programming matching

An on-line recognition system is presented for large-alphabet handprinted Chinese characters by the model based recognition approach with stroke based features. A deviation-expansion (D-E) model representing the reference pattern is constructed. The model contains the hypothetical knowledge of handwriting variations including stroke-order deviations and stroke-number deviations. For pattern matching a matching graph is constructed by combining the knowledge of the reference pattern and the unknown pattern together. With the graph a similarity measure function is defined to indicate similarity degree. The evaluation of the function is obtained by utilizing dynamic programming matching. Experimental results are based upon the testing set of 54,000 handprinted sample characters written in square style by ten persons. The D-E models of reference patterns saved in a data base are generated by 5400 daily-used Chinese characters. The unknown character to be recognized can be stroke-order and stroke-number free, tolerant for incorrect strokes and daily-used connected strokes, size and shape flexible. The cumulative classification rate of choosing the ten most similar characters is 98%. The results suggest that the hypothetical model is feasible and reasonable.