Assign Class Labels Research Articles

Inline inspection of textured plastics surfaces

This article focuses on the inspection of plastics web materials exhibiting irregular textures such as imitation wood or leather. They are produced in a continuous process at high speed. In this process, various defects occur sporadically. However, current inspection systems for plastics surfaces are able to inspect unstructured products or products with regular, i.e., highly periodic, textures, only. The proposed inspection algorithm uses the local binary pattern operator for texture feature extraction. For classification, semisupervised as well as supervised approaches are used. A simple concept for semisupervised classification is presented and applied for defect detection. The resulting defect-maps are presented to the operator. He assigns class labels that are used to train the supervised classifier in order to distinguish between different defect types. A concept for parallelization is presented allowing the efficient use of standard multicore processor PC hardware. Experiments with images of a typical product acquired in an industrial setting show a detection rate of 97% while achieving a false alarm rate below 1%. Real-time tests show that defects can be reliably detected even at haul-off speeds of 30 m/min. Further applications of the presented concept can be found in the inspection of other materials.

Learning aspect models with partially labeled data

In this paper, we address the problem of learning aspect models with partially labeled data for the task of document categorization. The motivation of this work is to take advantage of the amount of available unlabeled data together with the set of labeled examples to learn latent models whose structure and underlying hypotheses take more accurately into account the document generation process, compared to other mixture-based generative models. We present one semi-supervised variant of the Probabilistic Latent Semantic Analysis (PLSA) model ( Hofmann, 2001). In our approach, we try to capture the possible data mislabeling errors which occur during the training of our model. This is done by iteratively assigning class labels to unlabeled examples using the current aspect model and re-estimating the probabilities of the mislabeling errors. We perform experiments over the 20Newsgroups, WebKB and Reuters document collections, as well as over a real world dataset coming from a Business Group of Xerox and show the effectiveness of our approach compared to a semi-supervised version of Naive Bayes, another semi-supervised version of PLSA and to transductive Support Vector Machines.

PRED_PPI: a server for predicting protein-protein interactions based on sequence data with probability assignment

BackgroundProtein-protein interactions (PPIs) are crucial for almost all cellular processes, including metabolic cycles, DNA transcription and replication, and signaling cascades. Given the importance of PPIs, several methods have been developed to detect them. Since the experimental methods are time-consuming and expensive, developing computational methods for effectively identifying PPIs is of great practical significance.FindingsMost previous methods were developed for predicting PPIs in only one species, and do not account for probability estimations. In this work, a relatively comprehensive prediction system was developed, based on a support vector machine (SVM), for predicting PPIs in five organisms, specifically humans, yeast, Drosophila, Escherichia coli, and Caenorhabditis elegans. This PPI predictor includes the probability of its prediction in the output, so it can be used to assess the confidence of each SVM prediction by the probability assignment. Using a probability of 0.5 as the threshold for assigning class labels, the method had an average accuracy for detecting protein interactions of 90.67% for humans, 88.99% for yeast, 90.09% for Drosophila, 92.73% for E. coli, and 97.51% for C. elegans. Moreover, among the correctly predicted pairs, more than 80% were predicted with a high probability of ≥0.8, indicating that this tool could predict novel PPIs with high confidence.ConclusionsBased on this work, a web-based system, Pred_PPI, was constructed for predicting PPIs from the five organisms. Users can predict novel PPIs and obtain a probability value about the prediction using this tool. Pred_PPI is freely available at http://cic.scu.edu.cn/bioinformatics/predict_ppi/default.html.

Development of a microarray platform for FFPET profiling: application to the classification of human tumors

BackgroundmRNA profiling has become an important tool for developing and validating prognostic assays predictive of disease treatment response and outcome. Archives of annotated formalin-fixed paraffin-embedded tissues (FFPET) are available as a potential source for retrospective studies. Methods are needed to profile these FFPET samples that are linked to clinical outcomes to generate hypotheses that could lead to classifiers for clinical applications.MethodsWe developed a two-color microarray-based profiling platform by optimizing target amplification, experimental design, quality control, and microarray content and applied it to the profiling of FFPET samples. We profiled a set of 50 fresh frozen (FF) breast cancer samples and assigned class labels according to the signature and method by van 't Veer et al [1] and then profiled 50 matched FFPET samples to test how well the FFPET data predicted the class labels. We also compared the sorting power of classifiers derived from FFPET sample data with classifiers derived from data from matched FF samples.ResultsWhen a classifier developed with matched FF samples was applied to FFPET data to assign samples to either "good" or "poor" outcome class labels, the classifier was able to assign the FFPET samples to the correct class label with an average error rate = 12% to 16%, respectively, with an Odds Ratio = 36.4 to 60.4, respectively. A classifier derived from FFPET data was able to predict the class label in FFPET samples (leave-one-out cross validation) with an error rate of ~14% (p-value = 3.7 × 10-7). When applied to the matched FF samples, the FFPET-derived classifier was able to assign FF samples to the correct class labels with 96% accuracy. The single misclassification was attributed to poor sample quality, as measured by qPCR on total RNA, which emphasizes the need for sample quality control before profiling.ConclusionWe have optimized a platform for expression analyses and have shown that our profiling platform is able to accurately sort FFPET samples into class labels derived from FF classifiers. Furthermore, using this platform, a classifier derived from FFPET samples can reliably provide the same sorting power as a classifier derived from matched FF samples. We anticipate that these techniques could be used to generate hypotheses from archives of FFPET samples, and thus may lead to prognostic and predictive classifiers that could be used, for example, to segregate patients for clinical trial enrollment or to guide patient treatment.

Constructing a new fuzzy classifier based on hierarchical fuzzy entropy

In an earlier work, Lee et al. (Lee et al., 2001) presented a simple and fast fuzzy classifier that employed fuzzy entropy to evaluate pattern distribution information in a pattern space. In this paper, we extend his work to propose a new fuzzy classifier based on hierarchical fuzzy entropy (FC‐HFE). We retained the main parts of the original structure and modified some methods (e.g., methods for deciding the number of intervals in each dimension and for assigning class labels). In addition, the hierarchical fuzzy entropy is proposed for partitioning the decision region. The proposed FC‐HFE improves classification accuracy and overcomes some of the drawbacks in the Lee et al method (Lee et al., 2001). The simulation results show that the classification rate of the proposed FC‐HFE is better than earlier methods.

Machine learning: a review of classification and combining techniques

Supervised classification is one of the tasks most frequently carried out by so-called Intelligent Systems. Thus, a large number of techniques have been developed based on Artificial Intelligence (Logic-based techniques, Perceptron-based techniques) and Statistics (Bayesian Networks, Instance-based techniques). The goal of supervised learning is to build a concise model of the distribution of class labels in terms of predictor features. The resulting classifier is then used to assign class labels to the testing instances where the values of the predictor features are known, but the value of the class label is unknown. This paper describes various classification algorithms and the recent attempt for improving classification accuracy--ensembles of classifiers.

Efficiency Considerations for Vertical kNN Text Categorisation

The importance of text mining stems from the availability of huge volumes of text databases holding a wealth of valuable information that needs to be mined. Text mining is a coarse area encompassing many finer branches one of which is text categorisation or text classification. Text categorisation is the process of assigning class labels to documents based entirely on their textual contents where we are given a document d, and asked to find its subject matter or class label, Ci. In this paper, an optimised k-Nearest Neighbours classifier that uses discretisation, the P-tree technology, and dimensionality reduction to achieve a high degree of accuracy, space utilisation and time efficiency is proposed. One of the fundamental contributions of this work is that as new samples arrive, the proposed classifier can find the k nearest neighbours to the new sample from the training space without a single database scan.

Uncertainty and Confidence in Land Cover Classification Using a Hybrid Classifier Approach

Traditional methods of land cover classification and mapping are limited in providing spatial data on the uncertainty of map labels. In this paper, we present a hybrid classifier approach using Decision Tree (DT) and ARTMAP 1 neural network to providing confidence or uncertainty information via majority voting and other rules. The hybrid classifier is tested with AVHRR data to mapping land cover of North America. The two classifiers (DT and ARTMAP) tend to make predictive errors in different contexts. They show 68% agreement in classifying land cover of North America. A set of rules is developed to assign class labels for pixels where the two classifiers disagree. Levels of confidence in the hybrid classification derived from their individual voting (ARTMAP) and probability (DT) are used to assign confidence. The approach outlined in this paper produces two products—a hybrid classification map as well as a confidence map based on the two classification schemes. The hybrid approach seems suitable to tackle a variety of classification problems in remote sensing and may ultimately aid map users in making more informed decisions. Background

The naive credal classifier

Convex sets of probability distributions are also called credal sets. They generalize probability theory by relaxing the requirement that probability values be precise. Classification, i.e. assigning class labels to instances described by a set of attributes, is an important domain of application of Bayesian methods, where the naive Bayes classifier has a surprisingly good performance. This paper proposes a new method of classification which involves extending the naive Bayes classifier to credal sets. Exact and effective solution procedures for naive credal classification are derived, and the related dominance criteria are discussed. Credal classification appears as a new method, based on more realistic assumptions and in the direction of more reliable inferences.

Some approaches to optimal cluster labeling with applications to remote sensing

This paper presents some approaches for the problem of labeling clusters using information from a given set of labeled and unlabeled patterns. Assigning class labels to the clusters is formulated as finding the best label assignment over all possible label assignments with respect to a criterion. Labeling clusters is also viewed as obtaining probabilities of class labels to the clusters with the maximization of the likelihood function and probability of correct labeling as criteria. Closed form solutions are obtained for the probabilities of class labels to the clusters by maximizing a lower bound on the likelihood criterion. Fixed point iteration equations are developed for obtaining probabilities of class labels to the clusters. The problem of obtaining class labels to the clusters is further formulated as that of minimizing the variance of the proportion estimates of the classes that use both the given labeled and unlabeled patterns. Imperfections in the labels of the given labeled set are incorporated into the criteria. Furthermore, the results of application of these techniques in the processing of remotely sensed multispectral scanner imagery data are presented.