From its quirky start as a teenage iPhone chatbot to becoming the central hub for open‑source transformer models, Hugging Face has grown into a fast‑rising AI platform, securing $100 million Series C funding, serving thousands of organizations, and aiming to democratize machine learning.
K-means is a classic distance‑based clustering algorithm that iteratively partitions data into k compact, well‑separated groups by minimizing the sum of squared errors, using random centroid initialization and heuristic updates until convergence, making it a fundamental tool in AI and data analysis.
Systematic clustering, a widely used hierarchical clustering technique, builds a dendrogram by iteratively merging the closest sample points based on a distance matrix, allowing analysts to visualize and select groupings at various distance thresholds, from a single cluster to each point as its own class.
This article explains how to quantify similarity between data samples using distance metrics such as Manhattan, Euclidean, and Chebyshev, outlines the properties these distances must satisfy, and describes common inter‑class measures like single linkage, complete linkage, centroid, group average, and sum‑of‑squares methods.
Polynomial regression extends linear models by fitting data with higher‑order polynomial functions, requiring selection of the polynomial degree and its coefficients, and can be applied alongside other nonlinear fitting techniques to capture complex growth trends in real‑world systems.
This article introduces the basics of simple (univariate) linear regression and extends to multivariate linear regression, explaining their regression equations, the use of the least‑squares method to estimate parameters, and the practical relevance of multiple predictors in modeling real‑world phenomena.
PaSca, a scalable graph neural architecture search system that separates message aggregation from updates, explores over 150,000 GNN designs with multi‑objective optimization, delivers models that outperform traditional GNNs in accuracy, memory and speed, has been open‑sourced and deployed at Tencent for risk control, recommendation and fraud detection, and earned the WWW 2022 Best Student Paper award.
Harbor’s growth engine combines a passive, attribution‑driven traffic‑allocation algorithm with componentized ranking, search, and marketing systems—using pairwise/Listwise models, multi‑task CTR/CVR prediction, and automated strategy triggers—to align short‑term efficiency with long‑term LTV goals while moving toward causal inference and domain‑expert‑driven general models.
The GoodFuture International Algorithm Team, together with Jinan University Guangdong Smart Education Research Institute, distinguished themselves among 95 global teams in the 5th Educational Data Mining in Computer Science Education Workshop, securing a champion title in one task and a runner‑up in another, showcasing advanced AI‑driven predictive and recommendation techniques for intelligent student assessment.
Hello’s automated C‑side growth algorithm loop integrates diverse traffic sources, semi‑supervised PU‑learning, graph‑based look‑alike targeting, causal uplift models for smart subsidies, and adaptive copy and external ad optimization, dramatically boosting ride‑hailing and lifestyle service revenue while minimizing engineering duplication.
This report analyzes the metaverse’s technical foundation—covering edge computing, 5G/6G, machine learning, AIoT, holographic imaging, sensor tech, and decentralized cloud—while examining data growth in China, brain‑computer interaction policies, and the future of virtual avatars.
The Alibaba Mama technical team secured five paper acceptances at the prestigious KDD 2022 conference, presenting advances such as curriculum‑guided Bayesian reinforcement learning for ROI‑constrained bidding, adversarial‑gradient driven exploration for click‑through‑rate prediction, externality‑aware transformers for e‑commerce ads, multi‑modal multi‑query pretraining, and generative‑replay streaming graph neural networks.
This article details Amap's practical experience with Learning to Rank for search suggestions, covering application scenarios, data pipeline construction, feature engineering, model training, loss‑function adjustments, and the resulting performance improvements, while also discussing challenges such as sparse features and click bias.
Tulong is Meituan’s industrial graph neural network framework and learning platform that combines a compact MTGraph engine, a modular operator‑based GNN library, and visual workflow tools to enable heterogeneous, billion‑edge graph training on a single machine with up to 60 % memory savings and 2–4× speedups, streamlining search, recommendation, advertising and delivery pipelines.
This article explains how Tubi’s machine‑learning team reduces exposure bias in its video recommendation pipeline by normalizing popularity features, incorporating additional signals such as search behavior, and applying exploration techniques like bandit algorithms to diversify content exposure.
The 2021 Taobao New‑Product team boosted e‑commerce recommendation by redesigning the coarse‑ranking stage with a dual‑tower DSSM, low‑cost feature‑crossing, NOVA attention and multi‑task distillation from a fine‑ranking teacher, delivering up to +30‰ GAUC gain and 3‑5 % online CTR and click improvements.
This article walks through why accurate demand forecasting is critical for retailers, shows how to prepare and visualize sales data, demonstrates building a store‑item model with Databricks DDI and Facebook Prophet, and explains scaling the model to predict every product across all stores, highlighting performance metrics and practical tips.
This article introduces a curated list of notable PHP projects—including advanced machine‑learning libraries, a neural‑network framework, a natural‑language‑processing toolkit, a distributed long‑connection service, a database migration tool, a versatile filesystem abstraction, a C++ extension framework, and PHP‑FPM—highlighting their features, use‑cases, and sample code.
This article walks through implementing a basic multi‑layer neural network using only NumPy, covering terminology, network architecture, forward and backward propagation, activation functions, loss calculation, parameter updates with SGD, and compares the custom model with a Keras implementation.
This article presents an in‑depth overview of search term recommendation in QQ Browser, covering the various recommendation scenarios, the composition of recommendation items, the multi‑stage algorithm architecture, key technical challenges, evaluation metrics, and future research directions such as multi‑task and session‑aware modeling.
The article presents a machine‑learning‑assisted harmonic mean estimator that computes Bayesian model evidence without dependence on sampling strategies, explains its theoretical basis, compares it to the original estimator, and demonstrates its accuracy on Rosenbrock and Normal‑Gamma benchmarks.
PICASSO, Alibaba’s GPU‑centric sparse training engine for wide‑and‑deep recommender systems, merges identical embedding tables, interleaves data and kernel operations, and caches hot embeddings on GPU, eliminating the parameter server and delivering up to tenfold speedups over TensorFlow‑PS while maintaining model quality.
This article presents the DIRN model, which leverages product inclusion graphs and graph‑based embeddings to address the challenges of ranking both single‑item and complex travel products on Fliggy, demonstrating significant CTR, CVR, and GMV improvements through offline experiments and online A/B testing.
This article introduces regression analysis and explains why it is essential for predictive modeling, then details seven widely used regression techniques—including linear, logistic, polynomial, stepwise, ridge, lasso, and elastic‑net—while offering guidance on selecting the most appropriate model for a given dataset.
The article describes JD Search's machine‑learning alert system that combines offline and real‑time training, FFT‑based periodic detection, Prophet forecasting, and DBSCAN anomaly clustering, and explains architectural design, data preprocessing, model optimization, and distributed deployment to improve alert accuracy and response speed.
The article explains how normal (Gaussian) distributions underpin many machine‑learning algorithms, reviewing the central limit theorem, multivariate Gaussian sampling, and key properties such as products, sums, conditional and marginal distributions, linear transformations, and Gaussian‑based Bayesian inference.
This article details Baidu's large‑scale intelligent testing framework, describing how AST‑based unit test generation, automated API test creation, visual UI interaction case synthesis, GUI traversal action set generation, and front‑end assertion automation work together to achieve high‑coverage, low‑cost automated testing across multiple languages and platforms.
From the 1940s birth of computers to today's ultra‑large pre‑training models like Baidu’s ERNIE 3.0, AI has progressed through three development waves, now driven by algorithms, compute and data, with pre‑training lowering application barriers and evolving toward larger, multimodal, and more generalizable systems.
The article explains Shannon information content, entropy, cross‑entropy, KL‑divergence, conditional entropy and mutual information, illustrating each concept with coin‑flip and dice examples, visual formulas, and discusses their roles as loss functions and evaluation metrics in machine‑learning models.
NetEase Yanxuan built an end‑to‑end time‑series forecasting system for e‑commerce that integrates rich user, product, business and external features with a suite of statistical, machine‑learning and deep‑learning models, delivers predictions via a Tornado‑based service for thousands of SKUs, warehouses, advertising and app traffic, and shows that simpler models like XGBoost often outperform complex deep nets while interpretability and external shocks remain key challenges.
This presentation explores how to scientifically define, apply, evaluate, and interpret data in financial risk management, covering data alignment with business goals, feature selection, model metrics such as KS and PSI, handling pandemic effects, and methods for model explainability.
This article reviews the evolution of behavior‑sequence modeling methods—from pooling and target‑attention to RNN, capsule, transformer, and graph neural networks—explains their industrial relevance, and demonstrates how to quickly apply these techniques in the EasyRec framework with practical configuration examples.
This presentation explores how to scientifically define, apply, evaluate, and interpret data in financial risk management, covering data alignment with business goals, feature selection, model metrics like KS and PSI, handling pandemic impacts, and methods for model explanation and improvement.
The article describes JD Search’s machine‑learning‑based time‑series forecasting and anomaly‑detection platform, detailing its overall architecture, offline and real‑time training pipelines, FFT‑based periodicity detection, Prophet forecasting, DBSCAN outlier detection, and distributed optimizations such as Alink integration and load‑balancing strategies.
The article compares non‑intervention (natural) traffic, where models autonomously maximize UV, with intervention (allocation) traffic that fine‑tunes re‑ranking to meet short‑term business goals, outlines objectives of balancing immediate profit and long‑term value, and presents two technical solutions—an ML‑plus‑OR integer‑programming model and a PID‑based control loop—for real‑time traffic allocation.
The article describes how a logistics company built a data‑driven platform that uses big‑data storage, DeepFM and LSTM models, and real‑time GPS tracking to create an intelligent vehicle‑cargo matching system, a multi‑label driver tagging framework, and a freight price prediction engine, thereby improving efficiency and reducing costs across the industry.
This article introduces a practical, step‑by‑step tutorial on building machine learning models with scikit‑learn, covering problem types, dataset loading, splitting, and a series of reusable templates (V1.0, V2.0, V3.0) for classification, regression, clustering, cross‑validation, and hyper‑parameter tuning, complete with code examples.
The paper presents a generative re‑ranking framework for Taobao’s home‑decor channel that combines heuristic sequence generation methods (MMR, DPP, beam search) with a context‑aware encoder to produce diverse, relevance‑balanced recommendation lists, achieving notable gains in PV, IPV, CTR and click‑diversity over traditional point‑wise ranking.
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.
In this interview, Zuoyebang’s NLP lead explains how the company built self‑developed platforms like IQC and FTP to automate text quality inspection and intelligent labeling, outlines their architecture, shares practical deep‑learning applications such as translation and grammar correction, and discusses future research directions in large‑scale multi‑label classification, few‑shot learning, and multimodal models.
Time series forecasting, essential across weather, finance, and commerce, involves tasks like classification, clustering, anomaly detection, and especially prediction; this article explores its definitions, evaluation metrics, traditional methods, machine‑learning approaches, deep‑learning models such as TFT, and emerging AutoML tools, offering practical insights and best practices.
MDAP normalizes URLs by automatically learning pattern‑tree rule models using entropy‑based splits, gibberish and numeric detection, and scalable Flink processing, which groups millions of raw URLs into concise patterns for accurate statistical monitoring, dramatically reducing data noise while still facing latency and model‑iteration challenges.
This tutorial explains clustering as an unsupervised learning task, outlines why no single algorithm fits all data, and provides step‑by‑step Python code using scikit‑learn to install the library, generate synthetic datasets, and apply ten popular clustering algorithms with visualizations.
This talk explains how Ant Group's MarsStudio tackles the challenges of large‑scale marketing scenes by optimizing WebGL rendering to reduce CPU‑GPU communication and by using machine‑learning‑driven smart animation generation to let front‑end developers create dynamic effects quickly and efficiently.
Alibaba’s Taobao Live team replaced rule‑based bandwidth estimators with three machine‑learning solutions—Concerto, OnRL, and Loki—trained on over a million hours of global live‑stream data, achieving up to 13% throughput gain, threefold stall reduction, and up to 44% lower 95th‑percentile stalls, now deployed commercially.
The recall module in recommendation systems retrieves thousands of items from massive pools using parallel non-personalized and personalized paths—such as hot-item, content-based, behavior-based, and deep-model recall—prioritizing coverage and low latency while addressing challenges like hard-negative sampling, selection bias, objective alignment, and channel competition to feed downstream ranking.
This article explains what artificial intelligence is, distinguishes weak and strong AI, explores its applications in games, product testing, and NetEase's Fuxi platform, and introduces fundamental machine‑learning concepts such as supervised, unsupervised, and reinforcement learning, as well as neural networks and loss functions.
The article presents Snowball's model feature platform, detailing its motivation, architecture, feature lifecycle management, online engine design, optimization techniques, and the resulting improvements in feature iteration speed, reuse, and system stability for recommendation and search services.
This presentation details how Hallo Mobility tackles the challenges of information‑flow ad modeling by describing the ad ecosystem, the company’s business evolution, and the advertiser‑side algorithmic solutions—including plan‑level quality detection, creative‑level uplift modeling, feature‑cross engineering, and pre‑bid user screening—while outlining future directions for automated, data‑driven ad delivery.
This article describes how a logistics company built an intelligent, data‑driven platform that uses big‑data pipelines, DeepFM for cargo‑driver matching, LSTM for driver‑tag prediction, and cost‑aware models for freight‑price forecasting, improving efficiency and reducing operational costs.
The article surveys the re‑ranking stage of modern recommendation pipelines, detailing its architecture after recall and precise ranking, and examining how shuffling and diversity improve user experience, while multi‑task fusion, context‑aware learning‑to‑rank, real‑time online learning, and traffic‑control strategies balance accuracy, efficiency, and business responsiveness.
This article explains how Hello Mobility uses tree‑based causal inference and uplift modeling to improve smart subsidy efficiency in hotel marketing, covering background, uplift methods, custom split criteria, offline AUUC evaluation, online deployment, and future research directions.
This talk presents a comprehensive study of invalid query detection in voice assistants, covering the definition of effective and ineffective queries, challenges of non‑human interaction and ambiguous intent recognition, data collection, model design, experimental results, user‑feedback loops, and future research directions.
Hello Travel tackles information‑flow advertising challenges by using LightGBM‑based models to predict order conversion, creative performance, and pre‑bid user quality, augmenting sparse data with feature engineering and uplift techniques, while planning future fully automated delivery, richer pre‑screening, and cross‑channel reinforcement‑learning enhancements.
This article introduces Meituan's on‑site comprehensive knowledge graph, detailing its multi‑layer design, data‑driven construction pipeline, challenges of diverse user demands and industry complexity, and showcases practical applications in search, recommendation, intelligent display, as well as future expansion plans.
This article explores how combining causal inference with machine learning can uncover subtle correlations in large datasets, detailing user growth metrics, propensity‑score matching, causal recommendation models, heterogeneous treatment effect analysis, and practical strategies for improving retention and activity in recommendation systems.
This talk explains how Hello Travel tackles challenges in information‑flow advertising by describing the market landscape, their business background, and detailed algorithmic optimization across plan, creative, and pre‑bid dimensions, including data‑driven modeling, feature engineering, LightGBM and uplift models, and outlines future directions.
This article introduces the classic density‑based anomaly detection method Local Outlier Factor (LOF), explains its underlying concepts such as k‑distance, reachability distance, and local reachability density, outlines the algorithm steps, discusses its advantages and limitations, and provides practical Python examples using PyOD and scikit‑learn.
This article provides a comprehensive overview of machine learning, explaining its definition, relationship to AI, workflow, historical evolution, major paradigms, real‑world applications, and the most common algorithms with practical usage tips.
This article presents a comprehensive overview of 58.com’s real‑estate price system, describes how average prices are computed from platform data, explains three anomaly‑detection methods, and details a CatBoost‑based machine‑learning model for automated house valuation, including feature engineering and evaluation metrics.
This article presents Kuaishou's graph‑based item‑to‑item (I2I) recall pipeline for short‑video recommendation, detailing the business challenges, the advantages of graph neural networks, the system architecture, optimization techniques such as similarity metric refinement, graph structure learning, edge‑weight learning, and future research directions.
This talk presents a comprehensive study of invalid query detection in voice assistants, covering the definition and taxonomy of invalid queries, challenges of non‑human interaction and ambiguous intent recognition, data collection and labeling strategies, feature engineering, deep neural network modeling, experimental results, user‑feedback loops, and current performance limits.
This article presents Beike's experience in intelligent real‑estate advertising, detailing the business background, key technical challenges such as conversion‑rate estimation, delayed‑feedback modeling, GEO targeting, and budget allocation, and sharing practical solutions that improved conversion rates by over 10%.
This article details how QQ Music tackled rapid growth in recommendation traffic by redesigning its recommendation architecture, introducing a cloud‑native machine‑learning platform, optimizing data services, and adopting a DAG‑based recall system to improve scalability, flexibility, and development efficiency.
The article demonstrates how to build a digital twin for lithium‑ion batteries in Python by constructing a physics‑based model, augmenting it with experimental data using a simple Keras neural network, and visualizing predictions, illustrating the hybrid approach’s improved accuracy over purely empirical methods.
This paper introduces a novel multi-dimensional causal forest model combined with efficient integer programming algorithms to estimate heterogeneous treatment effects (HTE) in marketing scenarios, outperforming traditional tree-based methods through improved handling of intervention heterogeneity and resource allocation optimization.
The article introduces KuaiRec, a densely observed short‑video recommendation dataset with 99.6% density covering 1,411 users and 3,327 videos, discusses its structure, advantages over sparse public datasets, and its applicability to unbiased, interactive, conversational and reinforcement‑learning based recommendation studies.
This article presents a comprehensive overview of NetEase Cloud Music's advertising system, detailing its architecture, core challenges, CTR and CVR prediction models, feature engineering, model evolution from LR to deep learning, user vector modeling, and practical recommendations for improving ad performance.
This talk explores how causal inference techniques are used to identify and mitigate various biases in Kuaishou's recommendation pipeline, covering background theory, recent research advances, practical implementations for popularity and video completion debiasing, and reflections on challenges and future directions.
Fastbot, a collaborative AI‑powered testing service from ByteDance’s Quality Lab and GIP iOS platform team, overcomes TestFlight limits by using machine learning and reinforcement learning to automate stability testing, improve code coverage, detect accessibility issues, and streamline result consumption for faster app releases.
This article introduces the most important Python machine‑learning libraries of 2021—including TensorFlow, Scikit‑Learn, NumPy, Keras, PyTorch, LightGBM, Eli5, SciPy, Theano and Pandas—explaining their purposes, distinctive characteristics, and why they are essential tools for modern AI development.
The article introduces xeasy-ml, a Python machine‑learning framework, explains its installation requirements, demonstrates how to initialize a project, run training, view result artifacts, and provides detailed instructions for online prediction using configuration files and code examples.
This article examines how startups can harness massive pre‑trained AI models such as GPT‑3, outlining the historical context, benefits of transfer learning, the steep costs and data‑alignment challenges, and strategic considerations when using cloud APIs versus self‑hosting.
This article details Alibaba's exploration of intelligent supply‑chain forecasting, covering scenario classification, three generations of prediction algorithms, the self‑developed Falcon model, performance evaluation, and real‑world cases such as Double 11 and live‑streaming, highlighting challenges and practical solutions.
The paper applies causal inference and tree‑based uplift modeling to identify coupon‑responsive riders, using T‑, S‑, and X‑Learners as well as a proprietary Treelift model that directly optimizes per‑user utility, achieving a 4.7% lift over manual rules and 2.3% over prior response models.
This article presents a comprehensive overview of Taobao's user‑generated content (UGC) sentiment analysis pipeline, covering background, task definition, challenges, model architecture—including RoBERTa‑based extraction, sentiment‑knowledge pre‑training, and graph augmentation—personalized impression ranking, business impact cases, and future research directions.
This article introduces the Dxm Eros ultra‑large graph platform, explains its architecture, storage, analysis, modeling and visualization capabilities, and demonstrates how graph‑based machine learning is applied to fintech risk control, anti‑fraud, anti‑money‑laundering and automated audit workflows.
This article presents Beike's experience in applying intelligent advertising to the real‑estate industry, detailing business background, key technical challenges, multi‑task conversion modeling, GEO user targeting, delayed‑feedback calibration, and ROI‑driven budget allocation to improve ad effectiveness.
This article details JD's heuristic question‑answering framework for intelligent customer service, covering its pre‑consultation prediction, in‑consultation associative input, post‑consultation recommendation modules, underlying algorithms, deployment results, and future enhancement directions.
The article reviews three notable 2022 machine‑learning advancements—CatBoost, Amazon SageMaker's DeepAR forecasting, and the low‑code PyCaret library—explaining their unique advantages, typical use‑cases, and why they are expected to gain traction among data scientists.
This article describes how eBay applies graph neural networks to real‑time payment fraud detection, covering the anti‑fraud scenario, limitations of traditional GBDT pipelines, challenges of constructing and serving dynamic heterogeneous graphs, the end‑to‑end solution with directed slice graphs and a Lambda‑style architecture, and experimental results comparing GNN with LightGBM.
This talk explains how recommendation bias arises from popularity and position effects, introduces causal inference concepts and three inference levels, reviews recent research such as DICE and Huawei’s causal embedding, and details Kuaishou’s practical applications—including popularity debias, causal representation decoupling, and video completion‑rate debias—along with experimental results and future challenges.
This talk presents Alibaba's travel intent prediction system, detailing the unique challenges of low‑frequency, multi‑source travel behavior, the multi‑granular CNN and time‑attention model architecture, experimental comparisons with baselines, and how integrated user interest modeling improves recommendation performance.
This article presents an end‑to‑end solution that leverages graph neural networks and dynamic bipartite graph construction to detect payment fraud in eBay's e‑commerce platform in real time, addressing traditional model limitations, graph latency challenges, and demonstrating superior performance over GBDT approaches.
This article presents two straightforward music recommendation methods—content‑based filtering using audio features and collaborative filtering using playlist data—detailing their design ideas, key Python and Go code snippets, model training, evaluation, and possible improvements.
This article introduces the Dxm Eros ultra‑large graph platform, detailing its architecture, storage, analysis, modeling, and visualization modules, and demonstrates how graph machine‑learning techniques are applied to financial risk control, fraud detection, anti‑money‑laundering, and automated credit review.
This presentation explains the fundamentals of financial risk control, outlines the data and AI technologies needed across the pre‑loan, in‑loan, and post‑loan stages, and showcases visual‑analytics and federated‑learning case studies that improve model interpretability, monitoring, and enterprise risk management.
DeWu’s e‑commerce platform consolidated dozens of small‑scale transaction scenes into a universal personalized recommendation system by adopting a user‑to‑item DSSM dual‑tower model with unified sampling, category‑aware negative mining, cosine‑normalized embeddings, and real‑time serving, boosting click‑through rates by over 10% across all scenarios.
This article explores how combining causal inference with machine learning can detect subtle correlations in large datasets, improve user growth metrics such as retention and activity, and presents practical methods like propensity score matching, uplift modeling, HTE analysis, and meta‑learners applied to recommendation systems.
DeepMind’s AlphaCode, a transformer‑based AI trained on hundreds of gigabytes of code, demonstrated competitive programming abilities comparable to the middle 54% of human participants on Codeforces, highlighting significant progress in AI‑driven problem solving and code generation.
This article curates a wide range of high‑quality video tutorials, books, and online resources covering Python fundamentals, web crawling, data analysis, machine learning, Go language development, and front‑end technologies, offering practical recommendations and rating scores to help learners choose suitable study material.
This article explains why causal inference is essential for decision‑making, contrasts it with pure prediction, outlines the four DoWhy steps (modeling, identification, estimation, refutation), and demonstrates a case study on road freight accidents using DoWhy and EconML with code examples and results.
Debugging microservices in a Kubernetes environment is challenging due to the abstraction of pods, network complexities, infrastructure issues, and application-level faults, but by monitoring at the service layer, aggregating data, and applying machine‑learning‑based anomaly detection, teams can effectively identify and resolve problems.
This article presents Meituan's search ranking system, detailing its multi‑business sorting architecture, layered ranking pipeline, quota and fine‑ranking models, aggregation modeling techniques, and supporting platforms such as Lego and Poker, while also sharing practical insights and recruitment information.
The 2021 review of Alibaba’s Mama Search Advertising estimation platform details advances in model deepening—such as hash‑based embedding compression, adaptive dynamic parameters and graph neural networks—model interaction via a multi‑stage cascade with ranking distillation and oracle bias, and system efficiency gains from HPC training, mixed‑precision, multi‑hash embeddings, and fp16 quantization that deliver roughly a thirty‑fold speed‑up.
This article provides a comprehensive overview of fraud detection using graph neural networks, covering background definitions, fraud categories, GNN application steps, a timeline of key research papers, practical challenges, solutions, and a collection of open‑source resources and datasets.
This technical report presents a comprehensive overview of artificial intelligence evolution, knowledge‑graph construction techniques—including traditional, cross‑lingual and reading‑comprehension based named entity recognition, weak‑supervised and joint relation extraction, attribute completion via multi‑source cues, and conditional knowledge‑graph modeling—highlighting recent research findings and experimental results.
This article details the 58.com WPAI machine learning platform's architecture and the optimizations applied to training task scheduling, inference service elastic scaling, and offline‑online resource mixing, demonstrating how these techniques significantly improve GPU/CPU utilization and inference performance across both GPU and CPU environments.
The article reviews highlights from Flink Forward 2021, describing how real‑time computing is spreading across traditional industries, the unstoppable move toward Flink SQL, the emergence of Flink ML, and the vision of a streaming warehouse built on Flink Dynamic Table technology.
This article reviews the fundamentals of fraud, surveys the evolution of graph neural network research for fraud detection, outlines practical application steps, discusses key challenges such as disguise, scalability, and label scarcity, and provides representative papers, new research directions, industrial case studies, and open-source resources.
