A5101390996- Text and Document Classification Technologies
- Music and Audio Processing
- Advanced Text Analysis Techniques
- Semantic Web and Ontologies
- Spam and Phishing Detection
- MXene and MAX Phase Materials
- Imbalanced Data Classification Techniques
- Machine Learning and Data Classification
- Web Data Mining and Analysis
- Machine Learning in Bioinformatics
- Multimodal Machine Learning Applications
- Acute Myeloid Leukemia Research
- Sentiment Analysis and Opinion Mining
- Crime, Illicit Activities, and Governance
- Single-cell and spatial transcriptomics
- Advanced Computational Techniques and Applications
- Advanced Image and Video Retrieval Techniques
- Image Enhancement Techniques
- 2D Materials and Applications
- Domain Adaptation and Few-Shot Learning
- Multimedia Communication and Technology
- Environmental Monitoring and Data Management
- Remote Sensing and LiDAR Applications
- Graphene research and applications
- Machine Learning and ELM
Shanghai Jiao Tong University
2025
University of Toronto
2024
University of Electronic Science and Technology of China
2022-2023
Naval University of Engineering
2021
Southwest University
2017-2019
Southwestern University
2019
Huazhong University of Science and Technology
2016
Hebei University
2012
It is well-known that exploiting label correlations crucially important to multi-label learning. Most of the existing approaches take as prior knowledge, which may not correctly characterize real relationships among labels. Besides, are normally used regularize hypothesis space, while final predictions explicitly correlated. In this paper, we suggest for each individual label, prediction involves collaboration between its own and other Based on assumption, first propose a novel method learn...
Abstract Precise mapping of leukemic cells onto the known hematopoietic hierarchy is important for understanding cell-of-origin and mechanisms underlying disease initiation development. However, this task remains challenging because high interpatient intrapatient heterogeneity leukemia cell clones as well differences existed between normal cells. Using single-cell RNA sequencing (scRNA-seq) data with a curated clustering approach, we constructed comprehensive reference hematopoiesis. This...
Backdoor attacks have posed a serious threat in machine learning models, wherein adversaries can poison training samples with maliciously crafted triggers to compromise the victim model. Advanced backdoor attack methods focused on selectively poisoning more vulnerable samples, achieving higher success rate (ASR). However, we found that when manipulation strength of trigger is constrained very small value for imperceptible attacks, they suffer from extremely uneven class-wise ASR due unequal...
Partial multi-label learning (PML) deals with the problem where each training example is assigned multiple candidate labels, only a part of which are correct. To learn from such PML examples, straightforward model tends to be misled by noise label set. alleviate this problem, coupled framework established in paper desired and perform relabeling procedure alternatively. In procedure, instead simply extracting relative confidences, or deterministically eliminating low confidence labels...
Abstract Tuning electrical properties of 2D materials through mechanical strain has predominantly focused on n‐type like MoS 2 and WS , while p‐type such as WSe remain relatively unexplored. Here, the impact controlled electron transport characteristics both mono bi‐layer is studied. Through coupling atomic force microscopy (AFM) nanoindentation techniques conductive AFM, ability to finely tune electronic band structure demonstrated. The research offers valuable mechanistic insights into...
In multi-label learning, each instance is associated with multiple labels simultaneously. Most of the existing approaches directly treat label in a crisp manner, i.e. one class either relevant or irrelevant to instance. However, latent relative importance regrettably ignored. this paper, we propose novel learning approach that aims estimate labeling importances while training inductive model Specifically, present biconvex formulation both and graph regularization, solve problem using an...
This paper devotes the research on comparative reviews in field of mobiles. For reviews, study builds a feature lexicon and superlative lexicon, then classifies words according to their polarities. During extracting, an identifying method based naming characteristics is proposed for entities. results, constructs 6 extracting patterns apply reviews. Finally, formula given calculate result. The experimental results demonstrate feasibility this method.
Abstract In view of the deficiencies in traditional visual water surface object detection, such as existence non-detection zones, failure to acquire global information, and a single-shot multibox detector (SSD) detection algorithm remote low precision small objects, this study proposes from panoramic vision based on an improved SSD. We reconstruct backbone network for SSD algorithm, replace VVG16 with ResNet-50 network, add five layers feature extraction. More abundant semantic information...
Partial-label learning (PLL) solves the problem where each training instance is assigned a candidate label set, among which only one ground-truth label. The core of PLL to learn efficient feature representations facilitate disambiguation. However, existing methods plain by coarse supervision, incapable capturing sufficiently distinguishable representations, especially when confronted with knotty ambiguity, i.e., certain labels share similar visual patterns. In this paper, we propose novel...
Nowadays, many payment service providers use the discounts and other marketing strategies to promote their products. This also raises issue of people who deliberately take advantage such promotions reap financial benefits. These are known as ‘scalper parties’ or ‘econnoisseurs’ which can constitute an underground industry. In this paper, we show how machine learning assist in identifying abnormal scalper transactions. Moreover, introduce basic methods Decision Tree Boosting Tree, these...
Multimodal contrastive learning methods (e.g., CLIP) have shown impressive zero-shot classification performance due to their strong ability joint representation for visual and textual modalities. However, recent research revealed that multimodal on poisoned pre-training data with a small proportion of maliciously backdoored can induce CLIP could be attacked by inserted triggers in downstream tasks high success rate. To defend against backdoor attacks CLIP, existing defense focus either the...
Multimodal contrastive learning models (e.g., CLIP) can learn high-quality representations from large-scale image-text datasets, yet they exhibit significant vulnerabilities to backdoor attacks, raising serious safety concerns. In this paper, we disclose that CLIP's primarily stem its excessive encoding of class-irrelevant features, which compromise the model's visual feature resistivity input perturbations, making it more susceptible capturing trigger patterns inserted by attacks. Inspired...
In contrast to the standard learning paradigm where all classes can be observed in training data, with augmented (LAC) tackles problem unobserved data may emerge test phase. Previous research showed that given unlabeled an unbiased risk estimator (URE) derived, which minimized for LAC theoretical guarantees. However, this URE is only restricted specific type of one-versus-rest loss functions multi-class classification, making it not flexible enough when needs changed dataset practice. paper,...
In partial-label learning (PLL), each training example has a set of candidate labels, among which only one is the true label. Most existing PLL studies focus on instance-independent (II) case, where generation labels dependent However, this II-PLL paradigm could be unrealistic, since are usually generated according to specific features instance. Therefore, instance-dependent (ID-PLL) attracted increasing attention recently. Unfortunately, ID-PLL lack an insightful perception intrinsic...
The content in social media is difficult to analyze because of its informal and unstructured features. Luckily, some data like tweets have rich hashtags information, which can be helpful identify meaningful topic information. More importantly, the hashtag usually express context information a tweet best. To this end, paper introduces context-aware model detect track evolution by integrating time text-based media. Specifically, we develop two methods cope with different functions separately....
It is well-known that exploiting label correlations crucially important to multi-label learning. Most of the existing approaches take as prior knowledge, which may not correctly characterize real relationships among labels. Besides, are normally used regularize hypothesis space, while final predictions explicitly correlated. In this paper, we suggest for each individual label, prediction involves collaboration between its own and other Based on assumption, first propose a novel method learn...
MCC manipulation arising in recent years is a new kind of fraud happened financial system. Aiming at this issue, study proposes computational framework to detect such behavior. The utilizes hierarchical clustering discover the inherent transaction behavior pattern existed various businesses and predicts merchant by applying logistic regression model. experimental results indicate that model overall outperforms other classifiers detecting merchant's manipulation.
The dynamic nature of literature networks makes the task ranking scientific articles hard, hence we present a framework(IIRank) to determine influence proportion feature pairs in article on innovation and co-rank article. model is based title, keyword abstract information extracted from article, which make it possible consider pair as sensors their innovativeness importance, use Entropy method judge importance degree pair, greater degree, impact comprehensive evaluation.