Binary Task Research Articles

CTSVM: A robust twin support vector machine with correntropy-induced loss function for binary classification problems

As a variant of the support vector machine (SVM), the twin support vector machine (TSVM) has attracted substantial attention; however, TSVM is sensitive to outliers. To remedy it,this study introduces the correntropy-induced loss (C-loss) function, which is a non-convex, bounded, smooth loss function, and proposes the C-loss TSVM (CTSVM). CTSVM is designed to perform the following optimizations: minimization of the C-loss function, the 2-norm regularization of model coefficients, and the distance between the positive (negative) samples and the positive-class (negative-class) hyperplane, which ensures that negative (positive) samples are far away from the positive (negative) hyperplane. CTSVM is cast into two half-quadratic optimization problems that can be solved by using an alternating iterative method. This approach enables the computational time of CTSVM to remain comparable to that of related methods, especially when solving linear problems. Experimental results were processed statistically to compare the performance of multiple algorithms on multiple datasets. These results confirmed that CTSVM outperforms similar classifiers in terms of its robustness to outliers and classification accuracy for binary tasks. In the future, we aim to investigate the performance of CTSVM with respect to multi-class classification tasks.

How does positive mood modulate time-based event expectancy?

In the present study, we investigated how positive mood affects the formation of time-based event expectancies. After positive or neutral mood inductions, participants performed a binary choice response task in which two target stimuli (circle and square) and two pre-target intervals (800 and 1600ms) appeared equally often. One of the targets was paired with the short interval and the other target with the long interval in 90% of the trials. We found that participants from the positive and neutral groups showed markedly different behavioral patterns of time-based expectancy. The time-based expectancy was restricted to shorter intervals for the positive group and to longer intervals for the neutral group. We propose that positive mood increases attentional prioritization of information that is temporally closer to us.

Generalizing AUC Optimization to Multiclass Classification for Audio Segmentation With Limited Training Data

Area under the ROC curve (AUC) optimisation techniques developed for neural networks have recently demonstrated their capabilities in different audio and speech related tasks. However, due to its intrinsic nature, AUC optimisation has focused only on binary tasks so far. In this paper, we introduce an extension to the AUC optimisation framework so that it can be easily applied to an arbitrary number of classes, aiming to overcome the issues derived from training data limitations in deep learning solutions. Building upon the multiclass definitions of the AUC metric found in the literature, we define two new training objectives using a one-versus-one and a one-versus-rest approach. In order to demonstrate its potential, we apply them in an audio segmentation task with limited training data that aims to differentiate 3 classes: foreground music, background music and no music. Experimental results show that our proposal can improve the performance of audio segmentation systems significantly compared to traditional training criteria such as cross entropy.

Predicting molecular activity on nuclear receptors by multitask neural networks

AbstractThe interest in multitask and deep learning strategies has been increasing in the last few years, in application to large and complex dataset for quantitative structure‐activity relationship (QSAR) analysis. Multitask approaches allow the simultaneous prediction of molecular properties that are related, through information sharing, whereas deep learning strategies increase the potential of capturing nonlinear relationships. In this work, we compare the binary classification capability of multitask deep and shallow neural networks to single‐task strategies used as benchmark (i.e., as k‐nearest neighbours, N‐nearest neighbours, random forest and Naïve Bayes), as well as multitask supervised self‐organizing maps.Comparison was carried out with an extended QSAR dataset containing annotations of molecular binding, agonism and antagonism activity on 11 nuclear receptors, for a total of 14,963 molecules, divided into training and test sets and labelled for their bioactivity on at least one of 30 binary tasks. Additional 304 chemicals were used as external evaluation set to further validate models.Although no approach systematically overperformed the others, task‐specific differences were found, suggesting the benefit of multitask learning for tasks that are less represented. On average, some of the single‐task approaches and multitask deep learning strategies had similar performances. However, the latter can have advantages, such as a simpler management of predictions and applicability domain assessment for future samples. On the other hand, the parameter tuning required by neural networks are generally time expensive suggesting that the modelling strategy should be evaluated case by case.

On novelty detection for multi-class classification using non-linear metric learning

Novelty detection is a binary task aimed at identifying whether a test sample is novel or unusual compared to a previously observed training set. A typical approach is to consider distance as a criterion to detect such novelties. However, most previous work does not focus on finding an optimum distance for each particular problem. In this paper, we propose to detect novelties by exploiting non-linear distances learned from multi-class training data. For this purpose, we adopt a kernelization technique jointly with the Large Margin Nearest Neighbor (LMNN) metric learning algorithm. The optimum distance tries to keep each known class’ instances together while pushing instances from different known classes to remain reasonably distant. We propose a variant of the K-Nearest Neighbors (KNN) classifier that employs the learned distance to detect novelties. Besides, we use the learned distance to perform multi-class classification. We show quantitative and qualitative experiments conducted on synthetic and real data sets, revealing that the learned metrics are effective in improving novelty detection compared to other metrics. Our method also outperforms previous work regularly used for novelty detection.

Deep Mining External Imperfect Data for Chest X-Ray Disease Screening.

Deep learning approaches have demonstrated remarkable progress in automatic Chest X-ray analysis. The data-driven feature of deep models requires training data to cover a large distribution. Therefore, it is substantial to integrate knowledge from multiple datasets, especially for medical images. However, learning a disease classification model with extra Chest X-ray (CXR) data is yet challenging. Recent researches have demonstrated that performance bottleneck exists in joint training on different CXR datasets, and few made efforts to address the obstacle. In this paper, we argue that incorporating an external CXR dataset leads to imperfect training data, which raises the challenges. Specifically, the imperfect data is in two folds: domain discrepancy, as the image appearances vary across datasets; and label discrepancy, as different datasets are partially labeled. To this end, we formulate the multi-label thoracic disease classification problem as weighted independent binary tasks according to the categories. For common categories shared across domains, we adopt task-specific adversarial training to alleviate the feature differences. For categories existing in a single dataset, we present uncertainty-aware temporal ensembling of model predictions to mine the information from the missing labels further. In this way, our framework simultaneously models and tackles the domain and label discrepancies, enabling superior knowledge mining ability. We conduct extensive experiments on three datasets with more than 360,000 Chest X-ray images. Our method outperforms other competing models and sets state-of-the-art performance on the official NIH test set with 0.8349 AUC, demonstrating its effectiveness of utilizing the external dataset to improve the internal classification.

Boosting gender identification using author preference

• Virtual gender has never been utilized for authorship attribution in the literature. • We examine gender-swapping behavior in computer-mediated communications in detail. • We provide quantitative and empirical results for several research questions. • Gender classification accuracy improves via the integration of virtual gender. • We examine the imitation potential of different biological genders. Predicting the gender of a text document’s author, also known as gender identification, is a well-studied authorship categorization task in the literature. A common theme in gender identification studies is that gender is considered a binary task. However, digital communications provide users with the ability to select virtual genders leveraging physical anonymity. In this study, the additional duality on gender due to author preferences is examined along with the biological gender. Formally, the objective of this paper is to investigate whether the gender preference of an author contains any additional linguistic information. Furthermore, we explore whether this information can be exploited to improve the author characterization task. In particular, the self-assigned gender, i.e., virtual gender, of the users in text-based real-time online messaging services , along with the biological sex, is evaluated quantitatively via comparing/assessing the gender prediction performance under various settings. Experiment results show that by integrating the virtual gender into the binary classification problem of predicting an author’s gender, it is possible to further improve the prediction performance by 2.6%, up to 85.4%.

Assessing impact of channel selection on decoding of motor and cognitive imagery from MEG data

Objective. Magnetoencephalography (MEG) based brain–computer interface (BCI) involves a large number of sensors allowing better spatiotemporal resolution for assessing brain activity patterns. There have been many efforts to develop BCI using MEG with high accuracy, though an increase in the number of channels (NoC) means an increase in computational complexity. However, not all sensors necessarily contribute significantly to an increase in classification accuracy (CA) and specifically in the case of MEG-based BCI no channel selection methodology has been performed. Therefore, this study investigates the effect of channel selection on the performance of MEG-based BCI. Approach. MEG data were recorded for two sessions from 15 healthy participants performing motor imagery, cognitive imagery and a mixed imagery task pair using a unique paradigm. Performance of four state-of-the-art channel selection methods (i.e. Class-Correlation, ReliefF, Random Forest, and Infinite Latent Feature Selection were applied across six binary tasks in three different frequency bands) were evaluated in this study on two state-of-the-art features, i.e. bandpower and common spatial pattern (CSP). Main results. All four methods provided a statistically significant increase in CA compared to a baseline method using all gradiometer sensors, i.e. 204 channels with band-power features from alpha (8–12 Hz), beta (13–30 Hz), or broadband (α + β) (8–30 Hz). It is also observed that the alpha frequency band performed better than the beta and broadband frequency bands. The performance of the beta band gave the lowest CA compared with the other two bands. Channel selection improved accuracy irrespective of feature types. Moreover, all the methods reduced the NoC significantly, from 204 to a range of 1–25, using bandpower as a feature and from 15 to 105 for CSP. The optimal channel number also varied not only in each session but also for each participant. Reducing the NoC will help to decrease the computational cost and maintain numerical stability in cases of low trial numbers. Significance. The study showed significant improvement in performance of MEG-BCI with channel selection irrespective of feature type and hence can be successfully applied for BCI applications.

Technology of Informative Feature Selection for Immunosignature Analysis.

The main difficulty in practical work with data obtained via immunosignature analysis is high dimensionality and the presence of a significant number of uninformative or false-informative features due to the specific character of the technology. To ensure practically relevant quality of data analysis and classification, it is necessary to take due account of this specific character.The aim of the study is to create and test the technology for effective reduction of immunosignature data dimensionality, which provides practically relevant and high quality of classification with due regard for the properties of the data obtained.Materials and Methods.The study involved the use of two normalized data sets obtained from the public biomedical repository and containing the results of immunosignature analysis.The technology for selecting informative features was proposed within the framework of the study. It consisted of three successive steps: 1) breaking a multiclass task into a series of binary tasks using the “one vs all” strategy; 2) screening of false-informative features is performed for each binary comparison by comparing the values of the median of the sets “one” and “all”; 3) ranking of the remaining features according to their informative value and selection of the most informative ones for each binary comparison.To assess the quality of the proposed technology for informative feature selection, we used the results obtained after application of classification based on the filtered data. Support vector method that proved itself in the problems of high-dimensional data classification was used as a classification model.Results.Effectiveness of the proposed technology for informative feature selection was determined. This technology allows us to provide high quality of classification while significantly reducing the feature space. The number of features eliminated in the second step is approximately 50% for each data set under consideration, which greatly simplifies subsequent data analysis. After the third step, when the feature space is reduced to 15 features, the quality of classification by the macro-average F1-score metric is assessed as 98.9% for the GSE52581 dataset. For the GSE52581 dataset, with the feature space reduced to 266 features, the quality of classification by the macro-average F1-score metric is 91.3%.Conclusion.The results of the work demonstrate the promising outlook of the proposed technology for informative feature selection as applied to the data of immunosignature analysis.

Plea Bargaining: A Test of Dual Discounting Preferences for Non-Monetary Losses.

Judgment and decision-making research on discounting suggests that when humans are thinking about gains, they tend to prefer certain and immediate outcomes to uncertain and delayed outcomes. However, discounting has been studied primarily using monetary commodities and, until recently, by testing one feature of the binary forced-choice task at a time: delay, probability, or amount of money received/lost. The present research is the first test of a dual discounting task that combines probability and delay into a single, binary forced-choice task in a non-monetary loss context. The key findings, based on three studies, suggest that delay and probability discounting play a significant role in decisions including non-monetary loss commodities like plea bargaining. Future work should explore the boundary conditions of dual discounting based not only on the nature of the binary choice (probability and delay) but also on the nature of the commodity (amount, valence, and quantifiability).

Ironic efficiency in automation-aided signal detection

Decision makers often make poor use of the information provided by an automated signal detection aid; recent studies have found that participants assisted by an automated aid fell well short of best-possible sensitivity levels. The present study tested the generalisability of this finding over varying levels of aid reliability. Participants performed a binary signal detection task either unaided or with assistance from a decision aid that was 60%, 85%, or 96%-reliable. Assistance from a highly reliable aid (85% or 96%) improved discrimination performance, while assistance from a low-reliability aid (60%) did not. Because their ideal strategy is to place less weight on less reliable cues, however, the decision makers’ tendency to disuse the aid became more appropriate as the aid’s reliability declined. Automation-aided efficiency was thus near to optimal when the aid was close to chance but became highly inefficient, ironically, as the aid’s reliability increased. Practitioner Summary: Investigating operators’ automation-aided information integration strategies allows human factors practitioners to predict the level of performance the operator will attain. Ironically, in an aided signal detection task, performance when assisted by a highly reliable aid is far less efficient than that obtained when assisted by a far less reliable aid. Abbreviations: OW: optimal weighting; UW: uniform weighting; CC: contingent criterion; BD: best decides; CF: coin flip; PM: probability matching; HDI: highest density interval; MCMC: markov chain monte carlo; HR: hit rate; FAR: false alarm rate

Approximating the Ideal Observer for Joint Signal Detection and Localization Tasks by use of Supervised Learning Methods.

Medical imaging systems are commonly assessed and optimized by use of objective measures of image quality (IQ). The Ideal Observer (IO) performance has been advocated to provide a figure-of-merit for use in assessing and optimizing imaging systems because the IO sets an upper performance limit among all observers. When joint signal detection and localization tasks are considered, the IO that employs a modified generalized likelihood ratio test maximizes observer performance as characterized by the localization receiver operating characteristic (LROC) curve. Computations of likelihood ratios are analytically intractable in the majority of cases. Therefore, sampling-based methods that employ Markov-Chain Monte Carlo (MCMC) techniques have been developed to approximate the likelihood ratios. However, the applications of MCMC methods have been limited to relatively simple object models. Supervised learning-based methods that employ convolutional neural networks have been recently developed to approximate the IO for binary signal detection tasks. In this paper, the ability of supervised learning-based methods to approximate the IO for joint signal detection and localization tasks is explored. Both background-known-exactly and background-known-statistically signal detection and localization tasks are considered. The considered object models include a lumpy object model and a clustered lumpy model, and the considered measurement noise models include Laplacian noise, Gaussian noise, and mixed Poisson-Gaussian noise. The LROC curves produced by the supervised learning-based method are compared to those produced by the MCMC approach or analytical computation when feasible. The potential utility of the proposed method for computing objective measures of IQ for optimizing imaging system performance is explored.

Assisted deep learning framework for multi-class skin lesion classification considering a binary classification support

In this paper, we propose a deep convolutional neural network framework to classify dermoscopy images into seven classes. With taking the advantage that these classes can be merged into two (healthy/diseased) ones we can train a part of the network regarding this binary task only. Then, the confidences regarding the binary classification are used to tune the multi-class confidence values provided by the other part of the network, since the binary task can be solved more accurately. For both the classification tasks we used GoogLeNet Inception-v3, however, any CNN architectures could be applied for these purposes. The whole network is trained in the usual way, and as our experimental results on the skin lesion image classification show, the accuracy of the multi-class problem has been remarkably raised (by 7% considering the balanced multi-class accuracy) via embedding the more reliable binary classification outcomes.

Cascade architecture with rhetoric long short-term memory for complex sentence sentiment analysis

In a sentiment analysis task, it is essential to differentiate the various and sometimes even contradictory emotions in each segment and to judge the underlying true emotion of the whole sentence. Rhetorical structure theory hierarchically structures the relationships between segments and describes the effects of relations. This study proposes a flexible cascade architecture: the lower unit divides the sentence into segments and obtains their distributed representation vector; the upper rhetoric-based long short-term memory unit aggregates the information of every segment and applies the concrete hierarchical relation information to perform sentiment analysis. Auxiliary techniques, namely, data augmentation and relation clustering, are also proposed for preventing overfitting. The experiment results prove that the proposed cascade architecture and auxiliary techniques improve the traditional approaches in most cases, which shows 3.17% accuracy growth in the fine-grained classification and 1.41% in the binary tasks at most. Furthermore, the cascade architecture is flexible enough, which could be easily extended by combining the Rhetoric-LSTM unit with the state-of-the-art classification or pre-trained models if necessary.

Siamese neural networks for continuous disease severity evaluation and change detection in medical imaging

Using medical images to evaluate disease severity and change over time is a routine and important task in clinical decision making. Grading systems are often used, but are unreliable as domain experts disagree on disease severity category thresholds. These discrete categories also do not reflect the underlying continuous spectrum of disease severity. To address these issues, we developed a convolutional Siamese neural network approach to evaluate disease severity at single time points and change between longitudinal patient visits on a continuous spectrum. We demonstrate this in two medical imaging domains: retinopathy of prematurity (ROP) in retinal photographs and osteoarthritis in knee radiographs. Our patient cohorts consist of 4861 images from 870 patients in the Imaging and Informatics in Retinopathy of Prematurity (i-ROP) cohort study and 10,012 images from 3021 patients in the Multicenter Osteoarthritis Study (MOST), both of which feature longitudinal imaging data. Multiple expert clinician raters ranked 100 retinal images and 100 knee radiographs from excluded test sets for severity of ROP and osteoarthritis, respectively. The Siamese neural network output for each image in comparison to a pool of normal reference images correlates with disease severity rank (ρ = 0.87 for ROP and ρ = 0.89 for osteoarthritis), both within and between the clinical grading categories. Thus, this output can represent the continuous spectrum of disease severity at any single time point. The difference in these outputs can be used to show change over time. Alternatively, paired images from the same patient at two time points can be directly compared using the Siamese neural network, resulting in an additional continuous measure of change between images. Importantly, our approach does not require manual localization of the pathology of interest and requires only a binary label for training (same versus different). The location of disease and site of change detected by the algorithm can be visualized using an occlusion sensitivity map-based approach. For a longitudinal binary change detection task, our Siamese neural networks achieve test set receiving operator characteristic area under the curves (AUCs) of up to 0.90 in evaluating ROP or knee osteoarthritis change, depending on the change detection strategy. The overall performance on this binary task is similar compared to a conventional convolutional deep-neural network trained for multi-class classification. Our results demonstrate that convolutional Siamese neural networks can be a powerful tool for evaluating the continuous spectrum of disease severity and change in medical imaging.

Classification of glomerular hypercellularity using convolutional features and support vector machine.

Glomeruli are histological structures of the kidney cortex formed by interwoven blood capillaries, and are responsible for blood filtration. Glomerular lesions impair kidney filtration capability, leading to protein loss and metabolic waste retention. An example of lesion is the glomerular hypercellularity, which is characterized by an increase in the number of cell nuclei in different areas of the glomeruli. Glomerular hypercellularity is a frequent lesion present in different kidney diseases. Automatic detection of glomerular hypercellularity would accelerate the screening of scanned histological slides for the lesion, enhancing clinical diagnosis. Having this in mind, we propose a new approach for classification of hypercellularity in human kidney images. Our proposed method introduces a novel architecture of a convolutional neural network (CNN) along with a support vector machine, achieving near perfect average results on FIOCRUZ data set in a binary classification (lesion or normal). Additionally, classification of hypercellularity sub-lesions was also evaluated, considering mesangial, endocapilar and both lesions, reaching an average accuracy of 82%. Either in binary task or in the multi-classification one, our proposed method outperformed Xception, ResNet50 and InceptionV3 networks, as well as a traditional handcrafted-based method. To the best of our knowledge, this is the first study on deep learning over a data set of glomerular hypercellularity images of human kidney.

A Preliminary Study of Knowledge Transfer in Multi-Classification Using Gene Expression Programming.

Gene Expression Programming (GEP), a variant of Genetic Programming (GP), is a well established technique for automatic generation of computer programs. Due to the flexible representation, GEP has long been concerned as a classification algorithm for various applications. Whereas, GEP cannot be extended to multi-classification directly, and thus is only capable of treating an M-classification task as M separate binary classifications without considering the inter-relationship among classes. Consequently, GEP-based multi-classifier may suffer from output conflict of various class labels, and the underlying conflict can probably lead to the degraded performance in multi-classification. This paper employs evolutionary multitasking optimization paradigm in an existing GEP-based multi-classification framework, so as to alleviate the output conflict of each separate binary GEP classifier. Therefore, several knowledge transfer strategies are implemented to enable the interation among the population of each separate binary task. Experimental results on 10 high-dimensional datasets indicate that knowledge transfer among separate binary classifiers can enhance multi-classification performance within the same computational budget.

Reservoir computing with two-bit input task using dipole-coupled nanomagnet array

Reservoir computing (RC), which utilizes physical phenomena, is a candidate for low energy consumption recurrent neural networks. Recently, we proposed RC which uses a dipole-coupled nanomagnet array as a reservoir, which applies the direction of the static magnetic moment of each nanomagnet as a node state in the array. It is possible to learn binary tasks with one-bit input data using this reservoir. Moreover, more complex tasks such as motion detection utilize multi-bit input data. This study simulates a two-bit binary input RC by employing the dipole coupled nanomagnet array as a reservoir. We apply a binary task with two-bit input data, which requires a memory capacity and non-linear computing capability of the reservoir. Due to the simulations with a macrospin model at a temperature of 0 K, this RC was able to learn these binary tasks by utilizing 24 nanomagnets.

Detecting weak and strong Islamophobic hate speech on social media

ABSTRACTIslamophobic hate speech on social media is a growing concern in contemporary Western politics and society. It can inflict considerable harm on any victims who are targeted, create a sense of fear and exclusion amongst their communities, toxify public discourse and motivate other forms of extremist and hateful behavior. Accordingly, there is a pressing need for automated tools to detect and classify Islamophobic hate speech robustly and at scale, thereby enabling quantitative analyses of large textual datasets, such as those collected from social media. Previous research has mostly approached the automated detection of hate speech as a binary task. However, the varied nature of Islamophobia means that this is often inappropriate for both theoretically informed social science and effective monitoring of social media platforms. Drawing on in-depth conceptual work we build an automated software tool which distinguishes between non-Islamophobic, weak Islamophobic and strong Islamophobic content. Accuracy is 77.6% and balanced accuracy is 83%. Our tool enables future quantitative research into the drivers, spread, prevalence and effects of Islamophobic hate speech on social media.

Next-Generation Intermediate Representations for Binary Code Analysis

Many binary code analysis tools rely on intermediate representation (IR) derived from a binary code, instead of working directly with machine instructions. In this paper, we first consider binary code analysis problems that benefit from IR and compile a list of requirements that the IR suitable for solving these problems should meet. Generally speaking, a universal binary analysis platform requires two principal components. The first component is a retargetable instruction decoder that utilizes external specifications to describe target instruction sets. External specifications facilitate maintenance and allow one to quickly implement support for new instruction sets. We analyze some of the most popular instruction set architectures (ISAs), including those used in microcontrollers, and from that compile a list of requirements for the retargetable decoder. We then overview existing multi-ISA decoders and propose our vision of a more generic approach, based on a multi-layer directed acyclic graph that describes the decoding process in universal terms. The second component of the analysis platform is the actual architecture-neutral IR. In this paper, we describe such IRs and propose Pivot 2, an IR that is low-level enough to be easily constructed from decoded machine instructions, also being easy to analyze. The main features of Pivot 2 are explicit side effects, SSA variables, simpler alternative to phi-functions, and extensible elementary operation set at the core. This IR also supports machines that have multiple memory address spaces. Finally, we propose a way to tie the decoder and the IR together to fit them to most of the binary code analysis tasks through abstract interpretation on top of the IR. The proposed scheme takes into account various aspects of target architectures that are overlooked in many other works, including pipeline specifics (handling of delay slots, hardware loop support, etc.), exception and interrupt management, and generic address space model, in which accesses may have arbitrary side effects due to memory-mapped devices or other non-trivial behavior of the memory system.