Tagged articles
semi-supervised learning
42 articles · Page 1 of 1
ESA’s ESAC team introduced AnomalyMatch, a semi‑supervised active‑learning framework that, with fewer than ten labeled anomalies, processed roughly 100 million Hubble cutouts in just 2–3 days, uncovering 1,339 distinct anomalous astrophysical objects such as merging galaxies, gravitational lenses, and jellyfish galaxies.
Active LearningAnomalyMatchEfficientNet
0 likes · 16 min read


JD TechDec 30, 2025 · Artificial Intelligence
The paper introduces a semi‑supervised, extensible unified framework (SSUF) that integrates knowledge, label, and structural enhancements to overcome data sparsity, label bias, and fragmented sub‑tasks in e‑commerce query intent prediction, achieving superior offline and online performance.
BERTGCNe-commerce
0 likes · 14 min read


AntTechJul 11, 2024 · Information Security
The paper presents GIANTESS, a novel semi‑supervised fraud detection framework that leverages online‑identified suspicious transactions to augment the feature space, generating pseudo‑labels for out‑of‑distribution samples and employing a hybrid loss to improve detection of covert fraudulent activities, achieving notable recall gains on real‑world datasets.
GIANTESSmachine learningsemi-supervised learning
0 likes · 6 min read

Meituan's tech team highlights seven CVPR 2024 papers—spanning OCR pre‑training, long‑tail semi‑supervised learning, visual AIGC, audio‑visual segmentation and synthetic‑data detection—provides detailed abstracts and experimental results, and announces an online author‑talk session on June 27.
Audio-Visual SegmentationCVPR 2024Computer Vision
0 likes · 18 min read

This article explores how modern video AI techniques—ranging from transformer‑based classification to semi‑supervised retrieval and token‑halting acceleration—enable efficient, accurate detection of prohibited content and fast, scalable video search in the era of short‑form media.
AI moderationTransformersemi-supervised learning
0 likes · 18 min read


HomeTechJul 7, 2023 · Artificial Intelligence
This article presents a comprehensive multi-modal video understanding system for AIGC video generation, detailing technical architecture, GCN-based semi-supervised learning, and practical applications across automotive content scenarios.
AIGCBERTNeXtVLAD
0 likes · 8 min read

ARSL, an ambiguity‑resistant semi‑supervised learning framework for single‑stage object detection, introduces Joint‑Confidence Estimation and Task‑Separation Assignment to resolve selection and assignment ambiguities in pseudo‑labels, thereby markedly improving pseudo‑label quality and achieving state‑of‑the‑art AP gains on COCO benchmarks.
ARSLComputer VisionObject Detection
0 likes · 8 min read

PaddleDetection v2.6 expands the PP‑YOLOE family with rotating, small‑object, dense‑object, and ultra‑lightweight edge‑GPU models, upgrades PP‑Human and PP‑Vehicle toolboxes, releases semi‑supervised, few‑shot and distillation learning methods, adds numerous state‑of‑the‑art algorithms, and improves infrastructure with Python 3.10, EMA filtering and AdamW support.
BaiduComputer VisionFew-Shot Learning
0 likes · 14 min read

The article presents NetEase Yidun’s data‑centric AI approach to content moderation, covering the background of Data‑Centric AI, the specific business and data challenges of content safety, comprehensive data pipelines—including collection, labeling, augmentation, selection, cleaning, iteration and testing—and the role of self‑, semi‑ and weak‑supervised learning in enhancing algorithm performance.
Algorithm InnovationContent moderationData Management
0 likes · 19 min read


AntTechDec 19, 2022 · Artificial Intelligence
TransVCL introduces an end‑to‑end attention‑enhanced video copy localization network that leverages a custom Transformer, correlation‑Softmax similarity matrix, and temporal alignment module, combined with a semi‑supervised learning framework, achieving state‑of‑the‑art performance on VCSL and VCDB benchmarks.
AIAttentionTransformer
0 likes · 13 min read

At QCon 2022, NetEase’s AI experts detailed their end‑to‑end approach for digital content risk control, covering data acquisition, semi‑supervised training, dynamic inference, cost‑effective deployment, and future directions, highlighting how AI can boost efficiency while managing escalating operational expenses.
AIContent moderationCost Optimization
0 likes · 26 min read


AntTechNov 6, 2022 · Artificial Intelligence
This article surveys recent Ant Group research on explainable fraud detection, including constraint‑adaptive rule‑set learning (CRSL), meta‑path guided rule generation (MetaRule), biased sampling for imbalanced ranking, and a semi‑supervised data‑augmentation framework (SDAT) for tabular data, highlighting their motivations, methodologies, deployments, and experimental results.
Data AugmentationGraph Neural Networksconstraint adaptive
0 likes · 18 min read

This article reviews Alibaba DAMO‑ConvAI’s work on large‑scale conversational AI, comparing pretrained language and dialogue models, introducing the SPACE semi‑supervised knowledge‑injection framework and the Proton probing‑tuning method for extracting and applying model knowledge to downstream tasks.
Pretrained Dialogue ModelProbing Tuningknowledge injection
0 likes · 21 min read


58 TechSep 29, 2022 · Artificial Intelligence
58.com’s AI Lab presents a comprehensive overview of its end‑to‑end speech recognition system, detailing data collection, semi‑supervised training, Efficient Conformer architecture, model compression, and deployment strategies that together achieve high accuracy across diverse acoustic conditions and large‑scale production workloads.
AIEfficient ConformerEnd-to-End
0 likes · 19 min read


HomeTechSep 8, 2022 · Artificial Intelligence
This article presents a comprehensive overview of concept tag mining for recommendation systems, describing unsupervised pattern‑matching, semi‑supervised AutoPhase, and supervised NER approaches, analyzing their advantages and drawbacks, and offering practical solutions to tag duplication and quality issues.
NERNLPUnsupervised Learning
0 likes · 11 min read


Code DAOMay 19, 2022 · Artificial Intelligence
This article explains an end‑to‑end semi‑supervised training pipeline for Transformer‑based NLP models, detailing the unsupervised language‑model pre‑training, supervised fine‑tuning, and the internal architecture of embeddings, encoder layers, and downstream tasks such as text classification and NER.
BERTMasked Language ModelNLP
0 likes · 9 min read


Code DAOApr 22, 2022 · Artificial Intelligence
This article reviews Ladder Networks and the Γ‑Model for semi‑supervised learning, explains how they minimize supervised and unsupervised costs, presents experimental results on MNIST, CIFAR‑10 and SVHN, and details the Virtual Adversarial Training method with TensorFlow code.
CIFAR-10Ladder NetworksMNIST
0 likes · 9 min read

This article presents Hello Mobility’s precise marketing system, covering its business background, value, framework, algorithmic capabilities such as Pu‑Learning LookAlike modeling, TSA semi‑supervised learning, and Graph Embedding, as well as identified pain points, project impact, and future directions for scaling and automation.
AIgraph embeddinglookalike modeling
0 likes · 12 min read

This article explains how semi‑supervised pseudo‑label learning can dramatically improve model accuracy by using a tiny five‑line code snippet that generates pseudo‑labels for unlabeled data, retrains a second model, and avoids data leakage with a proper validation set.
AIdata labelingpseudo-labeling
0 likes · 4 min read

The Beike Real Estate NLP team secured first place in the CCIR Cup 2021 Intelligent Human‑Computer Interaction track by applying semi‑supervised and transfer learning techniques to small‑sample intent recognition and slot filling, and also presented the large‑scale Mandarin dialect speech benchmark KeSpeech at NeurIPS 2021.
AI competitionBERTNLP
0 likes · 5 min read

This article explores the rapid rise of AI‑generated video deepfakes, examines the four main manipulation techniques, discusses the inherent security risks, and presents NetEase Yidun’s comprehensive detection framework—including face‑detection‑based classification, semi‑supervised learning, feature fusion, and model distillation—to combat content‑security threats.
AI securityComputer Visiondeepfake detection
0 likes · 12 min read

Meituan’s NLP team leverages few‑shot learning—using data‑augmentation, semi‑supervised, ensemble/self‑training, and domain‑adaptation techniques—to cut annotation costs, achieving 1–2 percentage‑point accuracy gains on internal benchmarks and deploying high‑performing models for tasks such as topic classification, fake‑review detection, and sentiment analysis, while planning broader platform and model extensions.
Active LearningData AugmentationFew-Shot Learning
0 likes · 29 min read

The article, authored by a former Stanford PhD now at Zoom, forecasts that by 2030 speech recognition will rely heavily on semi‑supervised learning, on‑device models, richer representations, and personalization, while applications such as transcription services and voice assistants will evolve modestly.
AISpeech Recognitionfuture trends
0 likes · 7 min read

This article presents Didi's exploration of few‑shot learning, data‑augmentation, semi‑supervised self‑training and multi‑task learning techniques to address the scarcity of labeled samples in safety and governance scenarios, demonstrating practical solutions and performance gains across various risk‑detection tasks.
AIData AugmentationFew-Shot Learning
0 likes · 15 min read


Didi TechApr 20, 2021 · Artificial Intelligence
To overcome scarce labeled data for safety and governance, Didi combines few‑shot learning with systematic data augmentation, self‑training semi‑supervised labeling, and multi‑task neural architectures, cutting labeling costs and reducing log‑loss by over 20% while boosting ROC‑AUC and PR‑AUC across harassment detection, expense‑complaint, and route‑intercept use cases.
AI safetyData AugmentationDidi
0 likes · 15 min read

The Youku AI team won the Media AI Alibaba Entertainment Video Object Segmentation Challenge by enhancing the STM model with a spatial‑constrained memory reader, ASPP‑HRNet refinement, ResNeSt‑101 backbone, and a multi‑stage training pipeline, while also devising an unsupervised framework that combines DetectoRS detection, HRNet mask refinement, STM‑based association, and key‑frame optimization to achieve 95.5% test score on a large, richly annotated video dataset.
Computer VisionSpatial Memory NetworksUnsupervised VOS
0 likes · 13 min read

This weekly roundup highlights MariaDB’s new R2DBC connector for Java, China’s 4600‑km quantum communication network, JD’s 16 AAAI‑2021 papers, Intel’s RealSense ID facial recognition, JupyterLab 3.0 release, JD’s Smart Community 2.0, the OpenViDial multimodal dialogue dataset, and the FixMatch semi‑supervised learning breakthrough.
JupyterLabQuantum CommunicationR2DBC
0 likes · 8 min read

The article explains how building account relationship graphs and applying both traditional graph algorithms and modern graph neural networks—through community detection, anomaly detection, semi‑supervised and unsupervised learning, and dynamic graph techniques—can effectively identify and dismantle fraud groups in online services.
AIGraph Algorithmsdynamic graphs
0 likes · 11 min read


Didi TechNov 5, 2020 · Artificial Intelligence
DiDi’s self‑learning ASR platform lets non‑technical users upload business data, automatically train, test and deploy models with semi‑supervised learning, hot‑word updates and LSTM rescoring, creating a closed‑loop pipeline that boosted vehicle voice‑interaction accuracy from around 80 % to over 95 % within months.
AILanguage Modelacoustic model
0 likes · 14 min read

This article investigates the practical performance of Google’s 2019 Unsupervised Data Augmentation (UDA) framework on real‑world financial text classification tasks, detailing experiments with limited labeled data, domain‑out‑of‑distribution samples, noisy labels, and comparisons between BERT and lightweight TextCNN models.
BERTData AugmentationFinancial NLP
0 likes · 21 min read

This article describes Alibaba's AI system for automatically filtering nauseating product images on Taobao, covering challenges such as cold‑start, class imbalance, and diverse visual features, and detailing solutions like semi‑supervised learning, active learning, OHEM‑cascade, attention mechanisms, and the resulting business impact.
Active LearningAttention Mechanisme‑commerce AI
0 likes · 15 min read

This article presents a comprehensive front‑fusion recognition pipeline for high‑definition map static obstacle detection, detailing depth‑aware mapping, precise multi‑sensor calibration, point‑cloud registration, and semi‑supervised learning techniques that improve detection accuracy over traditional image‑only methods.
AIHD mapObject Detection
0 likes · 11 min read

This article explains how integrating depth data from LiDAR and stereo cameras with image‑based perception through front‑fusion algorithms reduces semantic errors, enhances static obstacle mapping, and enables semi‑supervised spatial annotation for high‑precision maps used in autonomous driving.
LiDARfront fusionhigh-precision mapping
0 likes · 11 min read

iQIYI introduces a semi‑supervised Unknown Identity Rejection (UIR) loss that uses massive unlabeled face images to push unknown samples away from known class centers, improving open‑set face‑recognition accuracy, feature sparsity, and out‑of‑library rejection rates across multiple benchmarks and products.
UIR lossface recognitionsemi-supervised learning
0 likes · 9 min read

Gaode’s engineering team presents a semi‑supervised deep‑learning framework that models road networks, extracts traffic, routing, deviation and heatmap features, and combines LSTM with ResNet to accurately identify dynamic road‑closure events, enabling both offline and real‑time detection with high confidence and business‑aligned validation.
LSTMResNetTraffic analysisbig data
0 likes · 12 min read


Amap TechSep 4, 2019 · Artificial Intelligence
Gaode’s semi‑supervised deep‑learning framework combines LSTM and ResNet to analyze 28‑day sequences of 39‑dimensional traffic and planning features, automatically discovering and verifying road‑closure events through a three‑layer pipeline, boosting detection confidence and top‑N accuracy by about ten percent and improving user routing.
LSTMResNetTraffic analysisroad closure detection
0 likes · 11 min read

The article explains how the heat diffusion equation and the Laplacian operator on graphs provide a physical intuition for Graph Convolutional Networks, showing the equivalence between continuous‑space Fourier analysis and discrete‑space message passing, and linking these concepts to semi‑supervised learning and GraphSAGE implementations.
GCNGraph Neural NetworksLaplacian
0 likes · 19 min read

The article explains the four main machine‑learning categories—supervised, unsupervised, semi‑supervised, and reinforcement learning—detailing how each approach handles data labeling, typical algorithms, and common application scenarios.
AI FundamentalsUnsupervised Learningmachine learning
0 likes · 2 min read

This article explains the distinction between discriminative and generative models, outlines the challenges of limited labeled data, introduces semi‑supervised learning principles, and describes GAN‑based semi‑supervised classification algorithms with illustrative diagrams.
GANGenerative Modelsartificial intelligence
0 likes · 3 min read

This article presents an Active PU Learning framework that combines active learning with two‑step PU semi‑supervised learning to improve cash‑out fraud detection, reducing labeling costs, enhancing model performance, and achieving a three‑fold increase in identified fraudulent transactions compared to traditional unsupervised methods.
AIActive Learningfraud detection
0 likes · 15 min read


AntTechApr 24, 2018 · Artificial Intelligence
This article summarizes a two‑stage method for anomaly detection when only a few labeled anomalies and many unlabeled instances are available, detailing problem formulation, isolation‑forest‑based scoring, clustering of anomalies, weighted multiclass modeling, experimental validation, and real‑world URL attack applications.
Pu-Learninganomaly detectionmachine learning
0 likes · 10 min read


AntTechApr 16, 2018 · Artificial Intelligence
This article presents an Active PU Learning framework that combines active learning with two‑step positive‑unlabeled learning to improve cash‑out fraud detection in Alipay’s fifth‑generation risk engine, AlphaRisk, achieving three‑fold identification gains over unsupervised methods while reducing labeling costs.
Active Learningfraud detectionmachine learning
0 likes · 14 min read
