Topic
Class (philosophy)
About: Class (philosophy) is a research topic. Over the lifetime, 821 publications have been published within this topic receiving 28000 citations.
Papers published on a yearly basis
Papers
More filters
TL;DR: In this article , Deepluenza, a deep learning model based on the BERT base multilingual model, was used to identify influenza reporting tweets. But the model was not applied to real-life influenza related data collected by health authorities.
Abstract: Influenza is a common seasonal disease that affects people worldwide. Quick reporting methods are needed to detect sudden influenza outbreaks so that health authorities can respond swiftly. Using social media posts to detect influenza related tweets may provide early insights about influenza outbreaks. In this paper, we introduce Deepluenza, a deep learning model to accurately identify influenza reporting tweets. Deepluenza supports multi-language (English and Arabic) Twitter streams. We conducted extensive experiments and compared the results obtained from Deepfluenza with real-life influenza related data collected by health authorities. Our experiments showed that Deepluenza, which is based on the BERT base multilingual model, achieved 0.99 accuracy and F1-score of 0.98 for the influenza reporting class, outperforming other conventional methods. The application of the developed model showed a positive correlation in the number of reports identified from social media with the number of actual hospital visits related to influenza. Furthermore, our experiments showed that combining tweets in different languages, such as English and Arabic, leads to an improved correlation between the number of posted tweets and the number of people’s visits to hospitals due to influenza infections. Deepluenza has the potential to be used for early detection of influenza outbreaks.
9 citations
TL;DR: In this article , the authors propose a novel approach to dynamically measure the instantaneous difficulty of each class during the training phase of the model, and then use the difficulty measures to design a novel weighted loss technique called "class-wise difficulty based weighted (CDB-W) loss" and a novel data sampling technique called ''class-wiscale difficulty based sampling'' for long-tailed datasets.
Abstract: Abstract Long-tailed datasets are very frequently encountered in real-world use cases where few classes or categories (known as majority or head classes) have higher number of data samples compared to the other classes (known as minority or tail classes). Training deep neural networks on such datasets gives results biased towards the head classes. So far, researchers have come up with multiple weighted loss and data re-sampling techniques in efforts to reduce the bias. However, most of such techniques assume that the tail classes are always the most difficult classes to learn and therefore need more weightage or attention. Here, we argue that the assumption might not always hold true. Therefore, we propose a novel approach to dynamically measure the instantaneous difficulty of each class during the training phase of the model. Further, we use the difficulty measures of each class to design a novel weighted loss technique called ‘class-wise difficulty based weighted (CDB-W) loss’ and a novel data sampling technique called ‘class-wise difficulty based sampling (CDB-S)’. To verify the wide-scale usability of our CDB methods, we conducted extensive experiments on multiple tasks such as image classification, object detection, instance segmentation and video-action classification. Results verified that CDB-W loss and CDB-S could achieve state-of-the-art results on many class-imbalanced datasets such as ImageNet-LT, LVIS and EGTEA, that resemble real-world use cases.
9 citations
TL;DR: In this article , two new classes of q-starlike functions in an open unit disc are defined and studied by using the q-fractional derivative, and the order of starlikeness in the class of convex functions is investigated.
Abstract: In this paper, two new classes of q-starlike functions in an open unit disc are defined and studied by using the q-fractional derivative. The class Sq*˜(α), α∈(−3,1], q∈(0,1) generalizes the class Sq* of q-starlike functions and the class Tq*˜(α), α∈[−1,1], q∈(0,1) comprises the q-starlike univalent functions with negative coefficients. Some basic properties and the behavior of the functions in these classes are examined. The order of starlikeness in the class of convex function is investigated. It provides some interesting connections of newly defined classes with known classes. The mapping property of these classes under the family of q-Bernardi integral operator and its radius of univalence are studied. Additionally, certain coefficient inequalities, the radius of q-convexity, growth and distortion theorem, the covering theorem and some applications of fractional q-calculus for these new classes are investigated, and some interesting special cases are also included.
9 citations
Posted Content•
TL;DR: In this paper, the convexity of the class of currents with finite relative energy was proved for relative non-pluripolar products, and the authors showed that the class is convex in the sense that all currents have the same relative energy.
Abstract: We prove the convexity of the class of currents with finite relative energy. A key ingredient is an integration by parts formula for relative non-pluripolar products which is of independent interest.
9 citations
9 citations