Elastic-net prefiltering for two-class classification

A two-stage linear-in-the-parameter model construction algorithm is proposed aimed at noisy two-class classification problems. The purpose of the first stage is to produce a prefiltered signal that is used as the desired output for the second stage which constructs a sparse linear-in-the-parameter classifier. The prefiltering stage is a two-level process aimed at maximizing a model's generalization capability, in which a new elastic-net model identification algorithm using singular value decomposition is employed at the lower level, and then, two regularization parameters are optimized using a particle-swarm-optimization algorithm at the upper level by minimizing the leave-one-out (LOO) misclassification rate. It is shown that the LOO misclassification rate based on the resultant prefiltered signal can be analytically computed without splitting the data set, and the associated computational cost is minimal due to orthogonality. The second stage of sparse classifier construction is based on orthogonal forward regression with the D-optimality algorithm. Extensive simulations of this approach for noisy data sets illustrate the competitiveness of this approach to classification of noisy data problems.

Incorporating label dependency into the binary relevance framework for multi-label classification

In multi-label classification, examples can be associated with multiple labels simultaneously. The task of learning from multi-label data can be addressed by methods that transform the multi-label classification problem into several single-label classification problems. The binary relevance approach is one of these methods, where the multi-label learning task is decomposed into several independent binary classification problems, one for each label in the set of labels, and the final labels for each example are determined by aggregating the predictions from all binary classifiers. However, this approach fails to consider any dependency among the labels. Aiming to accurately predict label combinations, in this paper we propose a simple approach that enables the binary classifiers to discover existing label dependency by themselves. An experimental study using decision trees, a kernel method as well as Naive Bayes as base-learning techniques shows the potential of the proposed approach to improve the multi-label classification performance.

Multi-parametric classification of Alzheimer's disease and mild cognitive impairment: the impact of quantitative magnetization transfer MR imaging

Multi-parametric and quantitative magnetic resonance imaging (MRI) techniques have come into the focus of interest, both as a research and diagnostic modality for the evaluation of patients suffering from mild cognitive decline and overt dementia. In this study we address the question, if disease related quantitative magnetization transfer effects (qMT) within the intra- and extracellular matrices of the hippocampus may aid in the differentiation between clinically diagnosed patients with Alzheimer disease (AD), patients with mild cognitive impairment (MCI) and healthy controls. We evaluated 22 patients with AD (n=12) and MCI (n=10) and 22 healthy elderly (n=12) and younger (n=10) controls with multi-parametric MRI. Neuropsychological testing was performed in patients and elderly controls (n=34). In order to quantify the qMT effects, the absorption spectrum was sampled at relevant off-resonance frequencies. The qMT-parameters were calculated according to a two-pool spin-bath model including the T1- and T2 relaxation parameters of the free pool, determined in separate experiments. Histograms (fixed bin-size) of the normalized qMT-parameter values (z-scores) within the anterior and posterior hippocampus (hippocampal head and body) were subjected to a fuzzy-c-means classification algorithm with downstreamed PCA projection. The within-cluster sums of point-to-centroid distances were used to examine the effects of qMT- and diffusion anisotropy parameters on the discrimination of healthy volunteers, patients with Alzheimer and MCIs. The qMT-parameters T2(r) (T2 of the restricted pool) and F (fractional pool size) differentiated between the three groups (control, MCI and AD) in the anterior hippocampus. In our cohort, the MT ratio, as proposed in previous reports, did not differentiate between MCI and AD or healthy controls and MCI, but between healthy controls and AD.

Expressive power of binary relevance and chain classifiers based on Bayesian Networks for multi-label classification

Bayesian network classifiers are widely used in machine learning because they intuitively represent causal relations. Multi-label classification problems require each instance to be assigned a subset of a defined set of h labels. This problem is equivalent to finding a multi-valued decision function that predicts a vector of h binary classes. In this paper we obtain the decision boundaries of two widely used Bayesian network approaches for building multi-label classifiers: Multi-label Bayesian network classifiers built using the binary relevance method and Bayesian network chain classifiers. We extend our previous single-label results to multi-label chain classifiers, and we prove that, as expected, chain classifiers provide a more expressive model than the binary relevance method.

Multi-dimensional classification of GABAergic interneurons with Bayesian network-modeled label uncertainty

Abstract Interneuron classification is an important and long-debated topic in neuroscience. A recent study provided a data set of digitally reconstructed interneurons classified by 42 leading neuroscientists according to a pragmatic classification scheme composed of five categorical variables, namely, of the interneuron type and four features of axonal morphology. From this data set we now learned a model which can classify interneurons, on the basis of their axonal morphometric parameters, into these five descriptive variables simultaneously. Because of differences in opinion among the neuroscientists, especially regarding neuronal type, for many interneurons we lacked a unique, agreed-upon classification, which we could use to guide model learning. Instead, we guided model learning with a probability distribution over the neuronal type and the axonal features, obtained, for each interneuron, from the neuroscientists’ classification choices. We conveniently encoded such probability distributions with Bayesian networks, calling them label Bayesian networks (LBNs), and developed a method to predict them. This method predicts an LBN by forming a probabilistic consensus among the LBNs of the interneurons most similar to the one being classified. We used 18 axonal morphometric parameters as predictor variables, 13 of which we introduce in this paper as quantitative counterparts to the categorical axonal features. We were able to accurately predict interneuronal LBNs. Furthermore, when extracting crisp (i.e., non-probabilistic) predictions from the predicted LBNs, our method outperformed related work on interneuron classification. Our results indicate that our method is adequate for multi-dimensional classification of interneurons with probabilistic labels. Moreover, the introduced morphometric parameters are good predictors of interneuron type and the four features of axonal morphology and thus may serve as objective counterparts to the subjective, categorical axonal features.

Multi-dimensional classification of GABAergic interneurons with Bayesian network-modeled label uncertainty

Interneuron classification is an important and long-debated topic in neuroscience. A recent study provided a data set of digitally reconstructed interneurons classified by 42 leading neuroscientists according to a pragmatic classification scheme composed of five categorical variables, namely, of the interneuron type and four features of axonal morphology. From this data set we now learned a model which can classify interneurons, on the basis of their axonal morphometric parameters, into these five descriptive variables simultaneously. Because of differences in opinion among the neuroscientists, especially regarding neuronal type, for many interneurons we lacked a unique, agreed-upon classification, which we could use to guide model learning. Instead, we guided model learning with a probability distribution over the neuronal type and the axonal features, obtained, for each interneuron, from the neuroscientists’ classification choices. We conveniently encoded such probability distributions with Bayesian networks, calling them label Bayesian networks (LBNs), and developed a method to predict them. This method predicts an LBN by forming a probabilistic consensus among the LBNs of the interneurons most similar to the one being classified. We used 18 axonal morphometric parameters as predictor variables, 13 of which we introduce in this paper as quantitative counterparts to the categorical axonal features. We were able to accurately predict interneuronal LBNs. Furthermore, when extracting crisp (i.e., non-probabilistic) predictions from the predicted LBNs, our method outperformed related work on interneuron classification. Our results indicate that our method is adequate for multi-dimensional classification of interneurons with probabilistic labels. Moreover, the introduced morphometric parameters are good predictors of interneuron type and the four features of axonal morphology and thus may serve as objective counterparts to the subjective, categorical axonal features.

Multi-class subcellular location prediction for bacterial proteins

Two algorithms, based onBayesian Networks (BNs), for bacterial subcellular location prediction, are explored in this paper: one predicts all locations for Gram+ bacteria and the other all locations for Gram- bacteria. Methods were evaluated using different numbers of residues (from the N-terminal 10 residues to the whole sequence) and residue representation (amino acid-composition, percentage amino acid-composition or normalised amino acid-composition). The accuracy of the best resulting BN was compared to PSORTB. The accuracy of this multi-location BN was roughly comparable to PSORTB; the difference in predictions is low, often less than 2%. The BN method thus represents both an important new avenue of methodological development for subcellular location prediction and a potentially value new tool of true utilitarian value for candidate subunit vaccine selection.

Machine Learning and Clinical Text. Supporting Health Information Flow

Fluent health information flow is critical for clinical decision-making. However, a considerable part of this information is free-form text and inabilities to utilize it create risks to patient safety and cost-­effective hospital administration. Methods for automated processing of clinical text are emerging. The aim in this doctoral dissertation is to study machine learning and clinical text in order to support health information flow.First, by analyzing the content of authentic patient records, the aim is to specify clinical needs in order to guide the development of machine learning applications.The contributions are a model of the ideal information flow,a model of the problems and challenges in reality, and a road map for the technology development. Second, by developing applications for practical cases,the aim is to concretize ways to support health information flow. Altogether five machine learning applications for three practical cases are described: The first two applications are binary classification and regression related to the practical case of topic labeling and relevance ranking.The third and fourth application are supervised and unsupervised multi-class classification for the practical case of topic segmentation and labeling.These four applications are tested with Finnish intensive care patient records.The fifth application is multi-label classification for the practical task of diagnosis coding. It is tested with English radiology reports.The performance of all these applications is promising. Third, the aim is to study how the quality of machine learning applications can be reliably evaluated.The associations between performance evaluation measures and methods are addressed,and a new hold-out method is introduced.This method contributes not only to processing time but also to the evaluation diversity and quality. The main conclusion is that developing machine learning applications for text requires interdisciplinary, international collaboration. Practical cases are very different, and hence the development must begin from genuine user needs and domain expertise. The technological expertise must cover linguistics,machine learning, and information systems. Finally, the methods must be evaluated both statistically and through authentic user-feedback.

Complex extreme learning machine applications in terahertz pulsed signals feature sets

This paper presents a novel approach to the automatic classification of very large data sets composed of terahertz pulse transient signals, highlighting their potential use in biochemical, biomedical, pharmaceutical and security applications. Two different types of THz spectra are considered in the classification process. Firstly a binary classification study of poly-A and poly-C ribonucleic acid samples is performed. This is then contrasted with a difficult multi-class classification problem of spectra from six different powder samples that although have fairly indistinguishable features in the optical spectrum, they also possess a few discernable spectral features in the terahertz part of the spectrum. Classification is performed using a complex-valued extreme learning machine algorithm that takes into account features in both the amplitude as well as the phase of the recorded spectra. Classification speed and accuracy are contrasted with that achieved using a support vector machine classifier. The study systematically compares the classifier performance achieved after adopting different Gaussian kernels when separating amplitude and phase signatures. The two signatures are presented as feature vectors for both training and testing purposes. The study confirms the utility of complex-valued extreme learning machine algorithms for classification of the very large data sets generated with current terahertz imaging spectrometers. The classifier can take into consideration heterogeneous layers within an object as would be required within a tomographic setting and is sufficiently robust to detect patterns hidden inside noisy terahertz data sets. The proposed study opens up the opportunity for the establishment of complex-valued extreme learning machine algorithms as new chemometric tools that will assist the wider proliferation of terahertz sensing technology for chemical sensing, quality control, security screening and clinic diagnosis. Furthermore, the proposed algorithm should also be very useful in other applications requiring the classification of very large datasets.

Standardized evaluation of algorithms for computer-aided diagnosis of dementia based on structural MRI: The CADDementia challenge

Algorithms for computer-aided diagnosis of dementia based on structural MRI have demonstrated high performance in the literature, but are difficult to compare as different data sets and methodology were used for evaluation. In addition, it is unclear how the algorithms would perform on previously unseen data, and thus, how they would perform in clinical practice when there is no real opportunity to adapt the algorithm to the data at hand. To address these comparability, generalizability and clinical applicability issues, we organized a grand challenge that aimed to objectively compare algorithms based on a clinically representative multi-center data set. Using clinical practice as the starting point, the goal was to reproduce the clinical diagnosis. Therefore, we evaluated algorithms for multi-class classification of three diagnostic groups: patients with probable Alzheimer's disease, patients with mild cognitive impairment and healthy controls. The diagnosis based on clinical criteria was used as reference standard, as it was the best available reference despite its known limitations. For evaluation, a previously unseen test set was used consisting of 354 T1-weighted MRI scans with the diagnoses blinded. Fifteen research teams participated with a total of 29 algorithms. The algorithms were trained on a small training set (n = 30) and optionally on data from other sources (e.g., the Alzheimer's Disease Neuroimaging Initiative, the Australian Imaging Biomarkers and Lifestyle flagship study of aging). The best performing algorithm yielded an accuracy of 63.0% and an area under the receiver-operating-characteristic curve (AUC) of 78.8%. In general, the best performances were achieved using feature extraction based on voxel-based morphometry or a combination of features that included volume, cortical thickness, shape and intensity. The challenge is open for new submissions via the web-based framework: http://caddementia.grand-challenge.org.

User Behavioral Modeling of Web-based Systems for Continuous User Authentication

Authentication plays an important role in how we interact with computers, mobile devices, the web, etc. The idea of authentication is to uniquely identify a user before granting access to system privileges. For example, in recent years more corporate information and applications have been accessible via the Internet and Intranet. Many employees are working from remote locations and need access to secure corporate files. During this time, it is possible for malicious or unauthorized users to gain access to the system. For this reason, it is logical to have some mechanism in place to detect whether the logged-in user is the same user in control of the user's session. Therefore, highly secure authentication methods must be used. We posit that each of us is unique in our use of computer systems. It is this uniqueness that is leveraged to "continuously authenticate users" while they use web software. To monitor user behavior, n-gram models are used to capture user interactions with web-based software. This statistical language model essentially captures sequences and sub-sequences of user actions, their orderings, and temporal relationships that make them unique by providing a model of how each user typically behaves. Users are then continuously monitored during software operations. Large deviations from "normal behavior" can possibly indicate malicious or unintended behavior. This approach is implemented in a system called Intruder Detector (ID) that models user actions as embodied in web logs generated in response to a user's actions. User identification through web logs is cost-effective and non-intrusive. We perform experiments on a large fielded system with web logs of approximately 4000 users. For these experiments, we use two classification techniques; binary and multi-class classification. We evaluate model-specific differences of user behavior based on coarse-grain (i.e., role) and fine-grain (i.e., individual) analysis. A specific set of metrics are used to provide valuable insight into how each model performs. Intruder Detector achieves accurate results when identifying legitimate users and user types. This tool is also able to detect outliers in role-based user behavior with optimal performance. In addition to web applications, this continuous monitoring technique can be used with other user-based systems such as mobile devices and the analysis of network traffic.

Hierarchical Multi-Label Text Classification in a Low-Resource Setting

In this thesis we address a multi-label hierarchical text classification problem in a low-resource setting and explore different approaches to identify the best one for our case. The goal is to train a model that classifies English school exercises according to a hierarchical taxonomy with few labeled data. The experiments made in this work employ different machine learning models and text representation techniques: CatBoost with tf-idf features, classifiers based on pre-trained models (mBERT, LASER), and SetFit, a framework for few-shot text classification. SetFit proved to be the most promising approach, achieving better performance when during training only a few labeled examples per class are available. However, this thesis does not consider all the hierarchical taxonomy, but only the first two levels: to address classification with the classes at the third level further experiments should be carried out, exploring methods for zero-shot text classification, data augmentation, and strategies to exploit the hierarchical structure of the taxonomy during training.

Land cover classification of Lago Grande de Curuai floodplain (Amazon, Brazil) using multi-sensor and image fusion techniques

Given the limitations of different types of remote sensing images, automated land-cover classifications of the Amazon várzea may yield poor accuracy indexes. One way to improve accuracy is through the combination of images from different sensors, by either image fusion or multi-sensor classifications. Therefore, the objective of this study was to determine which classification method is more efficient in improving land cover classification accuracies for the Amazon várzea and similar wetland environments - (a) synthetically fused optical and SAR images or (b) multi-sensor classification of paired SAR and optical images. Land cover classifications based on images from a single sensor (Landsat TM or Radarsat-2) are compared with multi-sensor and image fusion classifications. Object-based image analyses (OBIA) and the J.48 data-mining algorithm were used for automated classification, and classification accuracies were assessed using the kappa index of agreement and the recently proposed allocation and quantity disagreement measures. Overall, optical-based classifications had better accuracy than SAR-based classifications. Once both datasets were combined using the multi-sensor approach, there was a 2% decrease in allocation disagreement, as the method was able to overcome part of the limitations present in both images. Accuracy decreased when image fusion methods were used, however. We therefore concluded that the multi-sensor classification method is more appropriate for classifying land cover in the Amazon várzea.

Rgtsp: a generalized top scoring pairs package for class prediction.

SUMMARY: A top scoring pair (TSP) classifier consists of a pair of variables whose relative ordering can be used for accurately predicting the class label of a sample. This classification rule has the advantage of being easily interpretable and more robust against technical variations in data, as those due to different microarray platforms. Here we describe a parallel implementation of this classifier which significantly reduces the training time, and a number of extensions, including a multi-class approach, which has the potential of improving the classification performance. AVAILABILITY AND IMPLEMENTATION: Full C++ source code and R package Rgtsp are freely available from http://lausanne.isb-sib.ch/~vpopovic/research/. The implementation relies on existing OpenMP libraries.

Classification of urban multi-angular image sequences by aligning their manifolds

When dealing with multi-angular image sequences, problems of reflectance changes due either to illumination and acquisition geometry, or to interactions with the atmosphere, naturally arise. These phenomena interplay with the scene and lead to a modification of the measured radiance: for example, according to the angle of acquisition, tall objects may be seen from top or from the side and different light scatterings may affect the surfaces. This results in shifts in the acquired radiance, that make the problem of multi-angular classification harder and might lead to catastrophic results, since surfaces with the same reflectance return significantly different signals. In this paper, rather than performing atmospheric or bi-directional reflection distribution function (BRDF) correction, a non-linear manifold learning approach is used to align data structures. This method maximizes the similarity between the different acquisitions by deforming their manifold, thus enhancing the transferability of classification models among the images of the sequence.