Active Classification With Uncertainty Comparison Queries

Noisy pairwise comparison feedback has been incorporated to improve the overall query complexity of interactively learning binary classifiers. The positivity comparison oracle is extensively used to provide feedback on which is more likely to be positive in a pair of data points. Because it is impossible to determine accurate labels using this oracle alone without knowing the classification threshold, existing methods still rely on the traditional explicit labeling oracle, which explicitly answers the label given a data point. The current method conducts sorting on all data points and uses explicit labeling oracle to find the classification threshold. However, it has two drawbacks: (1) it needs unnecessary sorting for label inference and (2) it naively adapts quick sort to noisy feedback. In order to avoid these inefficiencies and acquire information of the classification threshold at the same time, we propose a new pairwise comparison oracle concerning uncertainties. This oracle answers which one has higher uncertainty given a pair of data points. We then propose an efficient adaptive labeling algorithm to take advantage of the proposed oracle. In addition, we address the situation where the labeling budget is insufficient compared to the data set size. Furthermore, we confirm the feasibility of the proposed oracle and the performance of the proposed algorithm theoretically and empirically.

Automatic Deep Learning-Based Consolidation/Collapse Classification in Lung Ultrasound Images for COVID-19 Induced Pneumonia

<div>Our automated deep learning-based approach identifies consolidation/collapse in LUS images to aid in the diagnosis of late stages of COVID-19 induced pneumonia, where consolidation/collapse is one of the possible associated pathologies. A common challenge in training such models is that annotating each frame of an ultrasound video requires high labelling effort. This effort in practice becomes prohibitive for large ultrasound datasets. To understand the impact of various degrees of labelling precision, we compare labelling strategies to train fully supervised models (frame-based method, higher labelling effort) and inaccurately supervised models (video-based methods, lower labelling effort), both of which yield binary predictions for LUS videos on a frame-by-frame level. We moreover introduce a novel sampled quaternary method which randomly samples only 10% of the LUS video frames and subsequently assigns (ordinal) categorical labels to all frames in the video based on the fraction of positively annotated samples. This method outperformed the inaccurately supervised video-based method of our previous work on pleural effusions. More surprisingly, this method outperformed the supervised frame-based approach with respect to metrics such as precision-recall area under curve (PR-AUC) and F1 score that are suitable for the class imbalance scenario of our dataset despite being a form of inaccurate learning. This may be due to the combination of a significantly smaller data set size compared to our previous work and the higher complexity of consolidation/collapse compared to pleural effusion, two factors which contribute to label noise and overfitting; specifically, we argue that our video-based method is more robust with respect to label noise and mitigates overfitting in a manner similar to label smoothing. Using clinical expert feedback, separate criteria were developed to exclude data from the training and test sets respectively for our ten-fold cross validation results, which resulted in a PR-AUC score of 73% and an accuracy of 89%. While the efficacy of our classifier using the sampled quaternary method must be verified on a larger consolidation/collapse dataset, when considering the complexity of the pathology, our proposed classifier using the sampled quaternary video-based method is clinically comparable with trained experts and improves over the video-based method of our previous work on pleural effusions.</div>

Automatic Deep Learning-Based Consolidation/Collapse Classification in Lung Ultrasound Images for COVID-19 Induced Pneumonia

<div>Our automated deep learning-based approach identifies consolidation/collapse in LUS images to aid in the diagnosis of late stages of COVID-19 induced pneumonia, where consolidation/collapse is one of the possible associated pathologies. A common challenge in training such models is that annotating each frame of an ultrasound video requires high labelling effort. This effort in practice becomes prohibitive for large ultrasound datasets. To understand the impact of various degrees of labelling precision, we compare labelling strategies to train fully supervised models (frame-based method, higher labelling effort) and inaccurately supervised models (video-based methods, lower labelling effort), both of which yield binary predictions for LUS videos on a frame-by-frame level. We moreover introduce a novel sampled quaternary method which randomly samples only 10% of the LUS video frames and subsequently assigns (ordinal) categorical labels to all frames in the video based on the fraction of positively annotated samples. This method outperformed the inaccurately supervised video-based method of our previous work on pleural effusions. More surprisingly, this method outperformed the supervised frame-based approach with respect to metrics such as precision-recall area under curve (PR-AUC) and F1 score that are suitable for the class imbalance scenario of our dataset despite being a form of inaccurate learning. This may be due to the combination of a significantly smaller data set size compared to our previous work and the higher complexity of consolidation/collapse compared to pleural effusion, two factors which contribute to label noise and overfitting; specifically, we argue that our video-based method is more robust with respect to label noise and mitigates overfitting in a manner similar to label smoothing. Using clinical expert feedback, separate criteria were developed to exclude data from the training and test sets respectively for our ten-fold cross validation results, which resulted in a PR-AUC score of 73% and an accuracy of 89%. While the efficacy of our classifier using the sampled quaternary method must be verified on a larger consolidation/collapse dataset, when considering the complexity of the pathology, our proposed classifier using the sampled quaternary video-based method is clinically comparable with trained experts and improves over the video-based method of our previous work on pleural effusions.</div>

A HYBRID SENTIMENT ANALYSIS APPROACH USING BLACK WIDOW OPTIMIZATION BASED FEATURE SELECTION

This paper proposes a novel hybrid framework with BWO based feature reduction technique which combines the merits of both machine learning and lexicon-based approaches to attain better scalability and accuracy. The scalability problem arises due to noisy, irrelevant and unique features present in the extracted features from proposed approach, which can be eliminated by adopting an effective feature reduction technique. In our proposed BWO approach, without changing the accuracy (90%), the feature-set size is reduced up to 43%. The proposed feature selection technique outperforms other commonly used PSO and GAbased feature selection techniques with reduced computation time of 21 sec. Moreover, our sentiment analysis approach is analysed using performance metrices such as precision, recall, F-measure, and computation time. Many organizations can use these online reviews to make well-informed decisions towards the users’ interests and preferences to enhance customer satisfaction, product quality and to find the aspects to improve the products, thereby to generate more profits.

Reliability Estimation of Inverse Lomax Distribution Using Extreme Ranked Set Sampling

In survival analysis, the two-parameter inverse Lomax distribution is an important lifetime distribution. In this study, the estimation of R = P Y < X is investigated when the stress and strength random variables are independent inverse Lomax distribution. Using the maximum likelihood approach, we obtain the R estimator via simple random sample (SRS), ranked set sampling (RSS), and extreme ranked set sampling (ERSS) methods. Four different estimators are developed under the ERSS framework. Two estimators are obtained when both strength and stress populations have the same set size. The two other estimators are obtained when both strength and stress distributions have dissimilar set sizes. Through a simulation experiment, the suggested estimates are compared to the corresponding under SRS. Also, the reliability estimates via ERSS method are compared to those under RSS scheme. It is found that the reliability estimate based on RSS and ERSS schemes is more efficient than the equivalent using SRS based on the same number of measured units. The reliability estimates based on RSS scheme are more appropriate than the others in most situations. For small even set size, the reliability estimate via ERSS scheme is more efficient than those under RSS and SRS. However, in a few cases, reliability estimates via ERSS method are more accurate than using RSS and SRS schemes.

DEPRESIVNOSTI VIZUELNO PAMĆENJE SPECIFIČNIH STIMULUSA (EMOTIKONA)

Depression is a state of reduced psychophysical activity that is accompanied by various changes in cognitive, emotional and social functioning. Previous studies have found that depression leads to changes in the recognition of the emotions of others, makes it difficult to direct attention and significantly impairs visual memory. The main goal of this research is to examine the relations between depressive symptoms and visual memory of specific stimuli that show emotions. We also want to examine whether the intensity of depressive symptoms is related to longer reaction time in the experimental task, as well as whether the accuracy of the emoticon’s visual memory depends on the set size. The research was conducted on a sample of 84 participants, students of Faculty of Philosophy in Banja Luka (90% female). The PHQ-9 questionnaire was used to assess depressive symptoms. Visual memory task was created in SuperLab 4.1. for Windows.The results show that there is a partial contribution of moderate depression to the accuracy of emoticon memory with sadness expression. No partial contribution of any category of depression to the memory accuracy of emoticons with the expression of happiness has been determined. A statistically significant negative correlation for the category of “sad” stimuli was obtained between the expression of depressive symptoms and the response time in the experimental task, while no statistically significant correlation was obtained for the second category of stimuli. It was also found that the number of errors increases with the set size. These results can be explained by negative bias and cognitive load in information processing. Key words: visual memory, depression, emoticons, expression of happiness, expression of sadness

Is categorization in visual working memory a way to reduce mental effort? A pupillometry study

Recent studies on visual working memory (VWM) have shown that visual information can be stored in VWM as continuous (e.g., a specific shade of red) as well as categorical representations (e.g., the general category red). It has been widely assumed, yet never directly tested, that continuous representations require more VWM mental effort than categorical representations; given limited VWM capacity, this would mean that fewer continuous, as compared to categorical, representations can be maintained simultaneously. We tested this assumption by measuring pupil size, as a proxy for mental effort, in a delayed estimation task. Participants memorized one to four ambiguous (boundaries between adjacent color categories) or prototypical colors to encourage continuous or categorical representations, respectively; after a delay, a probe indicated the location of the to-be-reported color. We found that, for set size 1, pupil size was larger while maintaining ambiguous as compared to prototypical colors, but without any difference in memory precision; this suggests that participants relied on an effortful continuous representation to maintain a single ambiguous color, thus resulting in pupil dilation while preserving precision. In contrast, for set size 2 and higher, pupil size was equally large while maintaining ambiguous and prototypical colors, but memory precision was now substantially reduced for ambiguous colors; this suggests that participants now also relied on categorical representations for ambiguous colors (which are by definition a poor fit to any category), thus reducing memory precision but not resulting in pupil dilation. Taken together, our results suggest that continuous representations are more effortful than categorical representations, and that very few continuous representations (perhaps only one) can be maintained simultaneously.

Unified Parameterization and Calibration of Serial, Parallel, and Hybrid Manipulators

In this work, we present methods allowing parallel, hybrid, and serial manipulators to be analyzed, calibrated, and controlled with the same analytical tools. We introduce a general approach to describe any robotic manipulator using established serial-link representations. We use this framework to generate analytical kinematic and calibration Jacobians for general manipulator constructions using null space constraints and extend the methods to hybrid manipulator types with complex geometry. We leverage the analytical Jacobians to develop detailed expressions for post-calibration pose uncertainties that are applied to describe the relationship between data set size and post-calibration uncertainty. We demonstrate the calibration of a hybrid manipulator assembled from high precision calibrated industrial components resulting in 91.1 μm RMS position error and 71.2 μrad RMS rotation error, representing a 46.7% reduction compared to the baseline calibration of assembly offsets.