Classification Paradigm Research Articles

CLIPMulti: Explore the performance of multimodal enhanced CLIP for zero-shot text classification

Zero-shot text classification does not require large amounts of labeled data and is designed to handle text classification tasks that lack annotated training data. Existing zero-shot text classification uses either a text–text matching paradigm or a text–image matching paradigm, which shows good performance on different benchmark datasets. However, the existing classification paradigms only consider a single modality for text matching, and little attention is paid to the help of multimodality for text classification. In order to incorporate multimodality into zero-shot text classification, we propose a multimodal enhanced CLIP framework (CLIPMulti), which employs a text–image&text matching paradigm to enhance the effectiveness of zero-shot text classification. Three different image and text combinations are tested for their effects on zero-shot text classification, and a matching method (Match-CLIPMulti) is further proposed to find the corresponding text based on the classified images automatically. We conducted experiments on seven publicly available zero-shot text classification datasets and achieved competitive performance. In addition, we analyzed the effect of different parameters on the Match-CLIPMulti experiments. We hope this work will bring more thoughts and explorations on multimodal fusion in language tasks.

Unveiling the Progenitors of a Population of Likely Peculiar Gamma-Ray Bursts

Traditionally, gamma-ray bursts (GRBs) are classified as long and short GRBs, with T 90 = 2 s being the threshold duration. Generally, long-duration GRBs (LGRBs; T 90 > 2 s) are associated with the collapse of massive stars, and short-duration (SGRBs; T 90 < 2 s) are associated with compact binary mergers involving at least one neutron star. However, the existence of a population of so-called “peculiar GRBs”—i.e., LGRBs originating from mergers or long Type I GRBs, and SGRBs originating from collapsars or short Type II GRBs—has challenged the traditional paradigm of GRB classification. Finding more peculiar GRBs may help to give more insight into this issue. In this work, we analyze the properties of machine-learning-identified long Type I GRB and short Type II GRB candidates, long GRBs-I and short GRBs-II (the so-called “peculiar GRBs”). We find that long GRBs-I almost always exhibit properties similar to Type I GRBs, which suggests that mergers may indeed produce GRBs with T 90 > 2 s. Furthermore, according to the probability given by the redshift distribution, short GRBs-II almost exhibit properties similar to Type II GRBs. This suggests that the populations of short Type II GRBs are not scarce and that they are hidden in a large number of samples without redshifts, which is unfavorable for the interpretation that the jet progression leads to a missed main emission.

Imbalanced graph learning via mixed entropy minimization

Imbalanced datasets, where the minority class is underrepresented, pose significant challenges for node classification in graph learning. Traditional methods often address this issue through synthetic oversampling techniques for the minority class, which can complicate the training process. To address these challenges, we introduce a novel training paradigm for node classification on imbalanced graphs, based on mixed entropy minimization (ME). Our proposed method, GraphME, offers a ‘free imbalance defense’ against class imbalance without requiring additional steps to improve classification performance. ME aims to achieve the same goal as cross-entropy-maximizing the model’s probability for the correct classes-while effectively reducing the impact of incorrect class probabilities through a “guidance” term that ensures a balanced trade-off. We validate the effectiveness of our approach through experiments on multiple datasets, where GraphME consistently outperforms the traditional cross-entropy objective, demonstrating enhanced robustness. Moreover, our method can be seamlessly integrated with various adversarial training techniques, leading to substantial improvements in robustness. Notably, GraphME enhances classification accuracy without compromising efficiency, a significant improvement over existing methods. The GraphME code is available at: https://github.com/12chen20/GraphME.

Understanding hate speech: the HateInsights dataset and model interpretability.

The persistence of hate speech continues to pose an obstacle in the realm of online social media. Despite the continuous evolution of advanced models for identifying hate speech, the critical dimensions of interpretability and explainability have not received proportional scholarly attention. In this article, we introduce the HateInsights dataset, a groundbreaking benchmark in the field of hate speech datasets, encompassing diverse aspects of this widespread issue. Within our dataset, each individual post undergoes thorough annotation from dual perspectives: firstly, conforming to the established 3-class classification paradigm that includes hate speech, offensive language, and normal discourse; secondly, incorporating rationales that outline specific segments of a post supporting the assigned label (categorized as hate speech, offensive language, or normal discourse). Our exploration yields a significant finding by harnessing cutting-edge state-of-the-art models: even models demonstrating exceptional proficiency in classification tasks yield suboptimal outcomes in crucial explainability metrics, such as model plausibility and faithfulness. Furthermore, our analysis underscores a promising revelation concerning models trained using human-annotated rationales. To facilitate scholarly progress in this realm, we have made both our dataset and codebase accessible to fellow researchers. This initiative aims to encourage collaborative involvement and inspire the advancement of the hate speech detection approach characterized by increased transparency, clarity, and fairness.

An Evaluation Methodology News Headline Classification using Comprehensive Pattern Learning

Text classification plays a pivotal role in numerous applications, ranging from spam detection to sentiment analysis. With the exponential growth of textual data stemming from social media interactions and news articles, the demand for accurate text classification methods has surged. This research delves into the evaluation of a novel weighting function designed to adjust the representation of texts during multilabel classification [1]. This innovative approach combines two distinct strategies: problem transformation and model adaptation. By incorporating this weighting function, the classification process undergoes a significant enhancement, ultimately improving the accuracy and efficiency of multilabel classification tasks [1]. To assess the efficacy of the proposed methodology, rigorous testing was conducted across ten referential textual datasets. Comparative analysis was performed against alternative techniques, with a keen focus on three key performance measures. The results obtained from these evaluations provide valuable insights into the effectiveness and robustness of the weighting function, shedding light on its potential to revolutionize text classification paradigms.

New clinical trials in CUP and a novel paradigm in cancer classification.

New clinical trials in CUP and a novel paradigm in cancer classification.

Eliminating quantization errors in classification-based sound source localization

Sound Source Localization (SSL) involves estimating the Direction of Arrival (DOA) of sound sources. Since the DOA estimation output space is continuous, regression might be more suitable for DOA, offering higher precision. However, in practice, classification often outperforms regression, exhibiting greater robustness. Conversely, classification’s drawback is inherent quantization error. Within the classification paradigm, the DOA output space is discretized into several intervals, each treated as a class. These classes show strong inter-class correlations, being inherently ordered, with higher similarity as intervals grow closer. Nevertheless, this characteristic has not been fully exploited. To address this, we propose Unbiased Label Distribution (ULD) to eliminate quantization error in training targets. Furthermore, we introduce Weighted Adjacent Decoding (WAD) to overcome quantization error during the decoding stage. Finally, we tailor two loss functions for the soft labels: Negative Log Absolute Error (NLAE) and Mean Squared Error without activation (MSE(wo)). Experimental results show our approach surpasses classification quantization limits, achieving state-of-the-art performance. Our code and supplementary material are available at https://github.com/linfeng-feng/ULD.

Multimodal self-supervised learning for remote sensing data land cover classification

Deep learning has revolutionized the remote sensing image processing techniques over the past few years. Nevertheless, annotating high-quality samples is difficult and time-consuming, which limits the performance of deep neural networks because of insufficient supervision information. Aiming to solve this contradiction, we investigate the multimodal self-supervised learning (MultiSSL) paradigm for pre-training and classification of remote sensing image. Specifically, the proposed self-supervised feature learning model consists of asymmetric encoder–decoder structure, in which deep unified encoder learns high-level key information characterizing multimodal remote sensing data and task-specific lightweight decoders are developed to reconstruct original data. To further enhance feature extraction capability, the cross-attention layers are utilized to exchange information contained in heterogeneous characteristics, thus learning more complementary information from multimodal remote sensing data. In fine-tuning stage, the pre-trained encoder and cross-attention layer serve as feature extractor, and leaned characteristics are combined with corresponding spectral information for land cover classification through a lightweight classifier. The self-supervised pre-training model can learn high-level key features from unlabeled samples, thereby utilizing the feature extraction capability of deep neural networks while reducing their dependence on annotated samples. Compared with existing classification paradigms, the proposed multimodal self-supervised pre-training and fine-tuning scheme achieves superior performance for remote sensing image land cover classification.

I-Biomarker CaDx: A circulating miRNA-based multi-cancer detection test with generative and explainable AI for colorectal cancer.

64 Background: Colorectal cancer (CRC), the second most deadly cancer, underscores the critical need for early detection to significantly improve treatment outcomes and survival rates. Colonoscopy, flexible sigmoidoscopy, and the fecal immunochemical tests (FIT), often fail to capture early disease, significantly decreasing the survival chance. Circulating miRNA is a promising non-invasive biomarker for various cancers, including CRC. The i-Biomarker CaDx (patent pending) is an innovative multi-cancer early detection and diagnosis platform, covering 32 cancer types with an impressive accuracy rate of 99-100%. This study focuses on the effectiveness of i-Biomarker CaDx for CRC, employing significant data collection and leveraging insights from Generative AI (GenAI) and Explainable Artificial Intelligence (XAI). Methods: We collected various available datasets (e.g. GSE106817, etc.) encompassing 305 colorectal cancer patients and a matching number of healthy controls. Microarray was employed to profile circulating miRNAs. We used various classification paradigms, e.g., decision trees, neural networks, etc. Hyperparameter optimization was conducted, and the best-performing classifiers were amalgamated in the final model, weighted by their respective efficacies. We assessed i-Biomarker CaDx's performance through cross-validation and independent test sets. The XAI provided detailed insights into miRNA variations associated with personalized diagnostic. GenAI integrated the test into the screening and early detection best practice workflows and performs Functional Analysis. Results: i-Biomarker CaDx shows exceptional diagnostic accuracy of 99-100%, surpassing conventional colorectal cancer diagnostic methods like colonoscopy, flexible sigmoidoscopy, and FIT. It can also be used for personalized treatment response monitoring. I-Biomarker effectively deciphers intricate miRNA relationships relevant to colorectal cancer using XAI. Thus, our analyses shed light on miRNA patterns and their associations with colorectal cancer, enhancing the understanding of the underlying molecular complexities of the disease. It goes further by explaining the personalized cellular and molecular mechanisms involved, and it will be integrated into the corresponding medical workflows. Conclusions: Our AI-powered multi-cancer early diagnostic platform demonstrates outstanding performance in detecting CRC, surpassing traditional diagnostic methods like colonoscopy, flexible sigmoidoscopy, and FIT. XAI allows for in-depth exploration of miRNA alterations and their impact on CRC, enriching our interpretations at both the population and individual levels. These findings underscore the substantial potential of i-Biomarker CaDx as a transformative, non-invasive diagnostic tool for CRC.

Multidimensional Emotional Disorder Inventory: reliability and validity in a Colombian non-clinical sample

BackgroundContemporary diagnostic frameworks in the realm of mental health have garnered criticism due to their categorical paradigm. Given the propensity of emotional disorders to manifest overlapping features, these frameworks fall short in comprehensively encapsulating their intricate nature. As a strategic response, Brown and Barlow introduced an innovative composite approach, amalgamating dimensions and categorical classifications, to adress this concern. Their strategic implementation hinged on the Multidimensional Emotional Disorder Inventory (MEDI), a transdiagnostic self-report instrument. Objective: this study undertakes the task of refining and validating the applicability of the MEDI within a non-clinical sample of Colombian university students (n = 808).MethodsThis study employed Exploratory Structural Equation Modeling (ESEM) to explore the structure of the measure. Results: ESEM suggested that the 8-dimension model with 48 items was the best-fitting solution, aligning with most dimensions identified by the original MEDI validation. Reliability was adequate for almost all dimensions (α: 0.69 – 0.92). An 8-dimension model with 48 items emerged as the most fitting solution, aligning with most dimensions identified by the original MEDI validation.ConclusionThe ensuing validation and contextual adaptation of the MEDI for use in the Colombian population augments the transdiagnostic evaluation of emotional disorders, with potential implications for enhanced stratification of targeted therapeutic interventions. By optimizing the assessment of both dimensional and cross-diagnostic paradigms, the MEDI portends a noteworthy impact in realms encompassing both academic inquiry and clinical practice.

Neural Network With a Preference Sampling Paradigm for Imbalanced Data Classification.

Most data in real life are characterized by imbalance problems. One of the classic models for dealing with imbalanced data is neural networks. However, the data imbalance problem often causes the neural network to display negative class preference behavior. Using an undersampling strategy to reconstruct a balanced dataset is one of the methods to alleviate the data imbalance problem. However, most existing undersampling methods focus more on the data or aim to preserve the overall structural characteristics of the negative class through potential energy estimation, while the problems of gradient inundation and insufficient empirical representation of positive samples have not been well considered. Therefore, a new paradigm for solving the data imbalance problem is proposed. Specifically, to solve the problem of gradient inundation, an informative undersampling strategy is derived from the performance degradation and used to restore the ability of neural networks to work under imbalanced data. In addition, to alleviate the problem of insufficient empirical representation of positive samples, a boundary expansion strategy with linear interpolation and the prediction consistency constraint is considered. We tested the proposed paradigm on 34 imbalanced datasets with imbalance ratios ranging from 16.90 to 100.14. The test results show that our paradigm obtained the best area under the receiver operating characteristic curve (AUC) on 26 datasets.

An interpretable deep learning Bayesian optimized random forest framework for the diagnosis of Parkinson's disease in structural magnetic resonance images

AbstractAs one of the fastest‐growing neurological disorders, Parkinson's disease is a neurodegenerative, cardinally motor ailment. Associated with the late onset of symptoms and often misdiagnosed, it inflicts massive life handicaps. To aid in its early detection, the research community is seeking alternative in‐vivo biomarkers, mainly in the field of neuroimaging. This study aimed to build a computer‐aided diagnosis system to distinguish between healthy controls (HC) and early Parkinsonian patients (PD). Our classification paradigm encompasses a Convolutional Neural Network built from scratch to extract discriminatory features, followed by a Bayesian‐optimized Random Forest classifier (CNN_RF). We trained and validated our customized model on the MRI images of 178 PD and 124 HC, taken from the benchmark Parkinson's Progression Markers Initiative (PPMI) dataset, as well as their brain tissue segments and Jacobian determinant maps. The classification was evaluated using seven performance metrics. For visual interpretation, we employed Grad‐CAM to generate saliency maps. Our findings demonstrate the robustness of the proposed method, as it outperformed deep‐tuned ResNet18, VGG16, and SqueezeNet, implemented for comparison. Furthermore, it resulted in 98.9% accuracy, 98.6% specificity, 99.0% precision, 99.5% F1‐score, a 97.7% Matthews correlation coefficient (MCC), and an AUC of 99.6% for the classification of fused gray matter and Jacobian map features. This indicates that the tailored CNN_RF possesses superior adaptability to identify intrinsic characteristics linked to the disease, as highlighted in the salient maps showcasing pertinent regions in the striatum, thalamus, frontal, temporal, and insular cortices for Parkinson's disease classification. Moreover, it offers a single modality and cost‐effective CAD, which could pave the way for a feasible early PD diagnosis. As data availability and AI robustness increase, PD diagnosis will no longer be a challenging task in the future.

Molecular subtypes of neuroendocrine carcinomas: A cross-tissue classification framework based on five transcriptional regulators

Neuroendocrine carcinomas (NECs) are extremely lethal malignancies that can arise at almost any anatomic site. Characterization of NECs is hindered by their rarity and significant inter- and intra-tissue heterogeneity. Herein, through an integrative analysis of over 1,000 NECs originating from 31 various tissues, we reveal theirtissue-independent convergence and further unveil molecular divergence driven by distinct transcriptional regulators. Pan-tissue NECs are therefore categorized into five intrinsic subtypes defined by ASCL1,NEUROD1, HNF4A, POU2F3, and YAP1. A comprehensive portrait of these subtypes is depicted, highlightingsubtype-specific transcriptional programs, genomic alterations, evolution trajectories, therapeutic vulnerabilities, and clinicopathological presentations. Notably, the newly discovered HNF4A-dominated subtype-H exhibits a gastrointestinal-like signature, wild-type RB1, unique neuroendocrine differentiation, poor chemotherapeutic response, and prevalent large-cell morphology. The proposal of uniform classification paradigm illuminates transcriptional basis of NEC heterogeneity and bridges the gap across different lineages and cytomorphological variants, in which context-dependent prevalence of subtypes underlies their phenotypic disparities.

Combined analysis of albumin in situ hybridisation and C reactive protein immunohistochemistry for the diagnosis of intrahepatic cholangiocarcinoma: towards a molecular classification paradigm

AimsIntrahepatic cholangiocarcinoma (iCCA) is a diagnosis of exclusion that can pose a challenge to the pathologist despite thorough clinical workup. Although several immunohistochemical markers have been proposed for iCCA, none...

Efficient identification of broad absorption line quasars using dimensionality reduction and machine learning

Efficient identification of broad absorption line quasars using dimensionality reduction and machine learning

Deep Representation Learning With Sample Generation and Augmented Attention Module for Imbalanced ECG Classification.

Developing an efficient heartbeat monitoring system has become a focal point in numerous healthcare applications. Specifically, in the last few years, heartbeat classification for arrhythmia detection has gained considerable interest from researchers. This paper presents a novel deep representation learning method for the efficient detection of arrhythmic beats. To mitigate the issues associated with the imbalanced data distribution, a novel re-sampling strategy is introduced. Unlike the existing oversampling methods, the proposed technique transforms majority-class samples into minority-class samples with a novel translation loss function. This approach assists the model in learning a more generalized representation of crucially important minority class samples. Moreover, by exploiting an auxiliary feature, an augmented attention module is designed that focuses on the most relevant and target-specific information. We adopted an inter-patient classification paradigm to evaluate the proposed method. The experimental results of this study on the MIT-BIH arrhythmia database clearly indicate that the proposed model with augmented attention mechanism and over-sampling strategy significantly learns a balanced deep representation and improves the classification performance of vital heartbeats.

Design of Transfer Learning based Deep CNN Paradigm for Brain Tumor Classification

Brain tumor is a fatal illness causing worldwide fatalities. The existing neuroimaging modalities to detect brain tumors are invasive and are observer-biased. Automatic CAD frameworks using sophisticated AI techniques lessen human intervention and can effectively handle large amounts of data. Automatic CAD frameworks using Machine learning techniques require the use of time-consuming and error-prone manual feature extraction procedures. Deep learning techniques involve automatic feature extraction; hence, appreciable classification results are attained quickly. However, training DL models from scratch takes a significant investment of time, money, and large datasets, which are difficult to attain in the medical domain. Therefore, the trade-off is utilizing the well exhaustively learned models like VGG16, VGG19, AlexNet, etc. to design a novel framework for the classification of brain tumors. The paper aims to develop a CNN-based deep learning framework by fine-tuning the pre-trained VGG16 architecture via transfer learning for brain tumor detection. The designed framework employing the transfer-learning technique gives better results with less data in less time. The brain tumor binary classification using brain MR images using transfer learning achieved an appreciable accuracy of 97%. The training and validation accuracy obtained was 100% and 97%, respectively, with 30 epochs. The loss for classification was as low as 0.0059% and the run time of 32ms/step time, much less than the existing models.

Fuzzy Graph Subspace Convolutional Network.

Graph convolutional networks (GCNs) are a popular approach to learn the feature embedding of graph-structured data, which has shown to be highly effective as well as efficient in performing node classification in an inductive way. However, with massive nongraph-organized data existing in application scenarios nowadays, it is critical to exploit the relationships behind the given groups of data, which makes better use of GCN and broadens the application field. In this article, we propose the f uzzy g raph s ubspace c onvolutional n etwork (FGSCN) to provide a brand-new paradigm for feature embedding and node classification with graph convolution (GC) when given an arbitrary collection of data. The FGSCN performs GC on the f uzzy s ubspace ( F -space), which simultaneously learns from the underlying subspace information in the low-dimensional space as well as its neighborliness information in the high-dimensional space. In particular, we construct the fuzzy homogenous graph GF on the F -space by fusing the homogenous graph of neighborliness GN and homogenous graph of subspace GS (defined by the affinity matrix of the low-rank representation). Here, it is proven that the GC on F -space will propagate both the local and global information through fuzzy set theory. We evaluated FGSCN on 15 unique datasets with different tasks (e.g., feature embedding, visual recognition, etc.). The experimental results showed that the proposed FGSCN has significant superiority compared with current state-of-the-art methods.

HyperFast: Instant Classification for Tabular Data

Training deep learning models and performing hyperparameter tuning can be computationally demanding and time-consuming. Meanwhile, traditional machine learning methods like gradient-boosting algorithms remain the preferred choice for most tabular data applications, while neural network alternatives require extensive hyperparameter tuning or work only in toy datasets under limited settings. In this paper, we introduce HyperFast, a meta-trained hypernetwork designed for instant classification of tabular data in a single forward pass. HyperFast generates a task-specific neural network tailored to an unseen dataset that can be directly used for classification inference, removing the need for training a model. We report extensive experiments with OpenML and genomic data, comparing HyperFast to competing tabular data neural networks, traditional ML methods, AutoML systems, and boosting machines. HyperFast shows highly competitive results, while being significantly faster. Additionally, our approach demonstrates robust adaptability across a variety of classification tasks with little to no fine-tuning, positioning HyperFast as a strong solution for numerous applications and rapid model deployment. HyperFast introduces a promising paradigm for fast classification, with the potential to substantially decrease the computational burden of deep learning. Our code, which offers a scikit-learn-like interface, along with the trained HyperFast model, can be found at https://github.com/AI-sandbox/HyperFast.

Cross-Modal Feature Distribution Calibration for Few-Shot Visual Question Answering

Few-shot Visual Question Answering (VQA) realizes few-shot cross-modal learning, which is an emerging and challenging task in computer vision. Currently, most of the few-shot VQA methods are confined to simply extending few-shot classification methods to cross-modal tasks while ignoring the spatial distribution properties of multimodal features and cross-modal information interaction. To address this problem, we propose a novel Cross-modal feature Distribution Calibration Inference Network (CDCIN) in this paper, where a new concept named visual information entropy is proposed to realize multimodal features distribution calibration by cross-modal information interaction for more effective few-shot VQA. Visual information entropy is a statistical variable that represents the spatial distribution of visual features guided by the question, which is aligned before and after the reasoning process to mitigate redundant information and improve multi-modal features by our proposed visual information entropy calibration module. To further enhance the inference ability of cross-modal features, we additionally propose a novel pre-training method, where the reasoning sub-network of CDCIN is pretrained on the base class in a VQA classification paradigm and fine-tuned on the few-shot VQA datasets. Extensive experiments demonstrate that our proposed CDCIN achieves excellent performance on few-shot VQA and outperforms state-of-the-art methods on three widely used benchmark datasets.