This article presents the 2020 Python library ranking, explaining the selection criteria and highlighting ten standout libraries—ranging from CLI tools like Typer and Rich to AI‑focused frameworks such as Hydra, PyTorch Lightning, Hummingbird, and HiPlot—plus several honorable mentions.
This article presents a comprehensive overview of a recommendation system, covering its business background, purpose, detailed engineering architecture—including data sources, computation, storage, online learning, service and access layers—and discusses key challenges, module design, and practical reflections.
This talk by Zhihu search algorithm engineer Shen Zhan details the evolution of text relevance models from TF‑IDF/BM25 to deep semantic matching and BERT, explains the challenges of deploying BERT at scale, and describes practical knowledge‑distillation techniques that improve both online latency and offline storage while maintaining search quality.
This article explains the fundamentals of recommendation systems, covering data collection, user and content profiling, system architecture, algorithmic pipelines such as recall, filtering, ranking, and evaluation metrics, while also discussing practical challenges like echo chambers and long‑term user value.
This article examines the technical evolution of POI expiration detection, covering attribute‑based, behavior‑based, and human‑place relationship mining methods, their machine‑learning models, and how they collectively improve map freshness and user experience at scale.
This article provides a comprehensive overview of e‑commerce recommendation systems, detailing their end‑to‑end workflow, key challenges such as multi‑scenario objectives and data loops, core components like recall and ranking, model evolution, feature engineering, evaluation metrics, and practical considerations for building a healthy, multi‑objective recommendation ecosystem.
This article introduces the concept of user preference mining, discusses challenges such as accurate expression, interpretability, and high-dimensional preferences, reviews statistical and model-based approaches including weighting, decay, XGBoost, DNN, LSTM, Seq4Rec, and Deep Interest Network, and describes their practical implementation at Beike.
DVC is an open‑source data version control system that extends Git to manage large machine‑learning models, datasets, and pipelines, enabling reproducible experiments, low‑friction branching, metric tracking, and seamless collaboration across various storage backends.
This article details the evolution of iQIYI's machine learning platform—from its early Javis‑based deep‑learning system to three major versions that introduced visual workflow, distributed scheduling, auto‑tuning, large‑scale training support, model management, and online prediction—while sharing practical lessons and a real anti‑cheat use case.
This article explains how to create a serverless CAPTCHA recognition service using a convolutional neural network (CNN) in Python, covering CAPTCHA types, data generation, model training, API implementation, front‑end integration, and deployment on Alibaba Cloud with detailed code examples.
This article introduces graph neural networks, explains their fundamentals and GraphSAGE/DGI algorithms, and demonstrates how Tencent applies them to recommendation scenarios such as video and WeChat content, highlighting network construction, feature engineering, sampling and aggregation techniques, and practical performance gains.
This article explains how Tencent's Angel platform powers game recommendation on Steam, Wegame, and the Tesla platform, comparing traditional linear tag‑based and collaborative‑filtering methods with the non‑linear DeepFM model and offering practical deployment tips.
The Wuhan‑Amap team won IPIN2020’s vehicle‑navigation track by using big‑data mining and neural‑network‑enhanced Vehicle‑Dead Reckoning to fuse smartphone GNSS, IMU, and barometer data, overcoming GPS outages and sensor limitations, and demonstrating that machine‑learning‑driven inertial navigation can achieve vehicle‑grade accuracy on consumer phones.
This article examines the challenges of moving machine‑learning models from offline training to online serving, proposes a modular architecture—including model gateway, data source gateway, business service center, monitoring, and RPC components—to enable rapid model deployment, version management, traffic mirroring, gray‑release, and real‑time monitoring.
This article explains the evolution of federated learning, contrasts Google’s cross‑device horizontal approach with China’s cross‑silo vertical implementations, analyzes their security vulnerabilities, and demonstrates how secure multi‑party computation—including differential privacy, secure aggregation, and secret‑sharing techniques—can address these challenges while highlighting performance trade‑offs.
This article explains how AI technologies are transitioning from labs to the web, covering neural network fundamentals, the distinction between cloud and edge intelligence, implementation pipelines, offline model optimization, online inference backends like WebGL and WASM, and practical web front‑end AI use cases.
This article explains the scale of online ad fraud, outlines common advertising billing models, describes how fake traffic generates revenue, defines invalid clicks, and presents a comprehensive anti‑fraud system that combines rule‑based methods, feature engineering, and AI models such as TextCNN, BiLSTM, BERT, Wide&Deep and GraphSage to identify and block fraudulent ad clicks.
This article explains the fundamentals of linear regression, including regression formulas, loss functions, and error metrics, and provides complete Python code using scikit‑learn to perform both simple and multiple linear regression on the Iris and Boston housing datasets, along with model evaluation and visualization.
This article explains the intuition behind gradient descent, describing how the cost function surface is visualized, how the learning rate controls step size, and why updating parameters in the correct order is essential for converging to the optimal solution.
The DataFunTalk Year-End Knowledge Graph Forum on December 19, 2023, will be streamed live and feature four expert speakers from Baidu, Alibaba, Meituan, and Beike who will share cutting‑edge knowledge‑graph technologies, applications, and practical techniques for industry and research audiences.
Learn how to translate common mathematical symbols such as summation, product, factorial, conditional expressions, and matrix multiplication into clear Python code, revealing the underlying computations and helping data scientists and ML practitioners deepen their mathematical intuition through practical examples.
This article shares Tubi's practical experience in rapidly iterating machine‑learning systems, emphasizing the early importance of simple end‑to‑end A/B testing platforms, clear launch plans, heat‑based and embedding‑based ranking models, and a culture of fast experimentation over complex deep‑learning research.
iQIYI’s Machine Learning Platform evolved from the specialized Javis deep‑learning system into a unified, low‑threshold solution for algorithm engineers, analysts, and developers, adding visual pipeline building, multi‑framework scheduling, automatic hyper‑parameter tuning, parameter‑server training, and scalable online prediction, dramatically boosting business efficiency and detection performance.
The article outlines a machine‑learning pipeline for automated captcha recognition, covering dataset generation, image preprocessing, segmentation via clustering or watershed methods, and classification using classic models and CNNs, achieving roughly 94% accuracy while noting the growing complexity of modern captchas and recommending developer collaboration when feasible.
This article examines the evolution of AI‑driven front‑end code generation—from early research like pix2code to Alibaba’s imgcook platform—detailing technical architectures, performance metrics, intelligent capability upgrades, and future directions for automated UI‑to‑code solutions.
The article shares Tubi's practical experience in building a fast‑iterating machine‑learning platform, emphasizing early measurement, simple end‑to‑end A/B testing, clear launch plans, lightweight popularity and embedding models, and rapid experimentation to drive product decisions.
This article examines the evolution of AI‑driven code generation for front‑end development, detailing the imgcook platform’s technical principles, performance metrics, intelligent capability upgrades, and its impact on development efficiency and workflow during Alibaba’s 2020 Double‑11 campaign.
This talk details the design, data pipeline, feature engineering, model evolution, and operational insights of the 58 Tongzhen home feed recommendation system, covering its architecture, localization strategies, recall and ranking models, online learning, and future directions for AI-driven content delivery in the down‑market.
The article reviews the importance of query intent recognition in vertical search, outlines its definition, highlights practical challenges such as ambiguous input, multi‑intent queries, timeliness and cold‑start issues, and surveys common rule‑based, statistical, and machine‑learning solutions together with real‑world case studies.
In this live talk, Dr. Zhang Renyu, an assistant professor at NYU Shanghai and economist at Kuaishou, presents his research on optimizing cold-start problems in feed advertising using a novel Shadow Bidding with Learning (SBL) algorithm, detailing its design, implementation, and experimental results.
This article details the design, data pipeline, feature engineering, model development, and iterative optimization of the 58 Tongzhen local feed recommendation system, covering business background, user profiling, recall strategies, ranking models such as XGBoost, XDeepFM, and online learning, and future directions.
Gaode’s TestPG platform solves full‑link load‑testing bottlenecks by unifying traffic capture with Iflow, converting logs into standardized corpora via a Flink pipeline, and applying corpus‑intelligence that extracts seasonal feature statistics and predicts distributions for precise, feature‑level throttling, enabling faster, more reliable testing and future autonomous optimization.
The article introduces the GEAR model, a graph‑based evidence aggregation and reasoning framework that leverages BERT representations and graph neural networks to improve multi‑evidence fact verification, discusses its challenges, experimental gains on the FEVER dataset, and potential applications such as fake‑news detection and knowledge‑graph validation.
Front‑end intelligentization leverages AI and machine learning to empower developers with design‑to‑code automation, smart UI generation, and Pipcook integration, aiming to solve frontline engineers' pain points, boost productivity, and evolve the front‑end technology stack through deterministic, robust, and evolutionary approaches.
This article explores the evolution of e‑commerce search, detailing why search matters, the technical pipeline—including query preprocessing, entity and intent recognition, knowledge‑graph construction, recall, coarse and fine ranking—and demonstrates substantial performance gains through real‑world case studies.
This article explains why front‑end intelligentization is needed, how AI and machine‑learning techniques are applied to design‑to‑code workflows, and the technical, operational, and cultural upgrades required to boost developer productivity and deliver business value in modern web projects.
A new MIT‑developed artificial‑intelligence system analyzes cough recordings captured on a phone or computer to identify COVID‑19 infections, achieving 98.5% accuracy and 100% detection of asymptomatic cases, offering a potential free, non‑invasive screening tool pending regulatory approval.
This article details the end‑to‑end evolution of 360's display advertising optimization, covering business flow, common ad formats, system architecture, CPC settings, traffic layering, smart bidding, creative combination, model progression from simple to deep learning, multi‑task learning, and latency reduction techniques.
This article presents a comprehensive overview of NetEase Yanxuan’s recall representation algorithms, detailing problem definition, model value, iterative implementations—including session-based embedding, GCN, GraphSAGE, LightGCN, and multi-interest models—along with engineering solutions, performance comparisons, and real-world deployment outcomes in search and recommendation systems.
The Hulu AI Class event showcased a series of technical talks covering large‑scale graph neural network optimizations, multi‑factor video ad placement algorithms, recommendation and search engine techniques, machine‑learning‑driven video codec improvements, and advanced content‑embedding methods, highlighting practical engineering experiences from Hulu’s Beijing office.
This article explains how Alibaba Cloud leverages AI and data‑driven automation to provide unattended, self‑service management for ECS instances, reducing operational costs, improving incident response speed, and ensuring stable, efficient cloud server operations.
This article traces the evolution of artificial neural networks from their biological inspiration, explains the McCulloch‑Pitts neuron model, details perceptron architecture and learning rule with a Scikit‑Learn example, and introduces multilayer perceptrons and the back‑propagation algorithm together with common activation functions.
Meituan’s AIOps initiative replaces manual rule‑based monitoring with the Horae platform, which automatically classifies time‑series metrics, applies CNN and XGBoost models to detect periodic anomalies, achieves over 90 % precision in production, and paves the way for broader metric types, forecasting, and advanced fault‑localization.
This article introduces ten essential algorithms—including Quick Sort, Heap Sort, Merge Sort, Binary Search, BFPRT, Depth‑First Search, Breadth‑First Search, Dijkstra's shortest‑path, Dynamic Programming, and Naive Bayes—explaining their principles, typical use cases, and step‑by‑step procedures.
The article explains cosine similarity from basic geometry and vector math, derives its formula, and shows how it powers precision marketing, image classification, and text retrieval, while also detailing its industrial implementation in Lucene’s vector space model.
This article introduces Baidu's new multi‑objective recall system (MOBIUS), which integrates relevance and business metrics such as CPM into the recall stage by migrating CTR models to recall, using data augmentation and a teacher‑student framework to improve ad monetization while preserving relevance.
This article describes how NetEase Yanxuan has designed, trained, and deployed a unified vector representation system to power various e‑commerce search and recommendation scenarios, covering model choices, incremental learning strategies, large‑scale similarity computation, and practical lessons from real‑world deployments.
This article introduces the Deep Sparse Network (NON), a new deep neural architecture for recommendation systems that combines field‑wise networks, across‑field interaction networks, and an operation‑fusion network, and demonstrates its superior performance through extensive experiments and ablation studies.
The 2020 Tencent Advertising Algorithm Live Week presented expert insights on federated learning, machine learning, big data, and deep‑learning applications in advertising, offering a comprehensive Q&A that explains how massive data fuels AI breakthroughs and reshapes business problem solving.
This weekly roundup covers a 20% surge in cloud service revenue, Google's new pQRNN model rivaling BERT, China's first NFC chip stamp, TSMC's 3nm capacity boost, Intel's IoT‑focused processors, Apple's open‑source Swift System for Linux, a novel FGPM adversarial text method, and multimodal translation advances, highlighting key trends across cloud, AI, and hardware.
The talk explores the current pain points of AI testing, outlines data‑quality analysis methods, highlights critical ETL and model‑testing considerations, and shares practical case studies and platform designs to improve machine‑learning quality assurance.
This article explores the definition of a model, distinguishes business, data, and function models, discusses criteria for a good model—including performance, fidelity to real‑world relationships, and interpretability—and examines why a universal model does not exist, all within the context of data science and AI.
This article introduces uplift (incremental) modeling as a causal inference technique for intelligent marketing, explains its mathematical formulation, compares response and uplift models, describes various modeling approaches such as two‑model, one‑model, and label‑transformation methods, outlines evaluation metrics like Qini and AUUC, and demonstrates practical deployment in a real‑world real‑estate platform.
Tencent Shield‑Federated Computing enables banks to jointly train Gradient Boosted Decision Trees and Logistic Regression with external data owners by using homomorphic encryption to perform encrypted variable and split‑point searches, gradient aggregation, and model updates, delivering near‑centralized accuracy, up to 70 % speed gains, and full data confidentiality for financial risk control.
This article explains how Yanxuan built a comprehensive vector system—from product embeddings and graph models to large‑scale similarity computation—and applied it across search, recommendation, and purchase prediction tasks, highlighting practical algorithms, infrastructure, and future directions.
Leveraging large-scale courier trajectory data and spatiotemporal analytics, the DTInf framework infers parcel delivery times by detecting stay points, correcting delivery locations, and matching delivery events using a trained MLP model, achieving a mean absolute error of 401 seconds and outperforming baselines by over 30%.
This article summarizes the Q&A session of the 2020 Tencent Advertising Algorithm Competition live series, covering the fundamentals of automated machine learning, its key technologies, current challenges, and the features and advantages of the SolnML system, while also addressing practical concerns such as hardware support and future research directions.
This article explains how to implement a user‑based collaborative filtering recommendation system in Python, covering data loading, preprocessing, cosine‑similarity computation, neighbor selection, rating prediction, and generating top‑5 movie recommendations with detailed code examples.
This article explains how keyword matching in search advertising works, outlines the challenges of semantic gaps, matching‑mode determination and scalability, and describes data‑driven synonym transformation techniques—including rule‑based, sequence‑to‑sequence, metric‑space and graph‑based models—to improve recall, efficiency, and robustness.
To enhance object detection robustness, the article discusses various data augmentation methods—including rotation, flipping, random cropping, scaling, color jitter, blurring, transparency adjustment, and image partitioning—providing code examples and illustrating their impact on model performance with before‑and‑after results.
This article provides an extensive overview of feature engineering, covering feature understanding, cleaning, construction, selection, transformation, and dimensionality reduction techniques, illustrated with Python code using the Titanic dataset, and offers practical guidelines for improving data quality and model performance in machine learning projects.
This article presents a comprehensive technical solution for a distributed online inference system that packages machine‑learning models in Docker containers, orchestrates them with Kubernetes for fault‑tolerant, elastic scaling, and integrates model repositories, image registries, monitoring, and automated model selection to streamline deployment, updates, and resource management.
This article summarizes the 2020 Tencent Advertising Algorithm Competition live‑stream series, presenting expert explanations of federated learning, its technical background, typical use cases, privacy and security mechanisms, and answers to common questions about applying federated learning to digital marketing.
This article presents Alibaba's latest ranking innovations for directed e‑commerce advertising, detailing the challenges of long‑term user interest modeling, the Search‑Based Interest Model (SIM) that extends behavior sequences to ten thousand actions, and the Dynamic Computation Allocation Framework (DCAF) that optimizes per‑request compute resources to maximize system revenue.
This article summarizes recent tech news, covering TikTok's AI‑driven recommendation system, U.S. pressure on ByteDance, Microsoft's migration of Visual Studio Codespaces to GitHub Codespaces, Debian's developer shortage despite strong funding, and the new features introduced in Laravel 8.
This article shares ten practical Python‑focused machine‑learning engineering tips—from writing abstract classes and fixing random seeds to tracking progress, speeding up pandas, managing cloud costs, and building robust FastAPI services—helping developers write cleaner, more reproducible, and production‑ready code.
This article walks through loading the Speed Dating dataset, examining its features and missing values, visualizing match rates by gender and age, performing correlation analysis, and building a logistic regression model with SMOTE oversampling to predict whether a pair will successfully match.
This comprehensive tutorial introduces machine learning fundamentals, its history, differences from traditional programming, key characteristics, and why Python is the preferred language, then explores supervised, unsupervised, and reinforcement learning, popular algorithms, detailed K‑Nearest Neighbors examples for classification and regression, and the essential steps to build and evaluate models.
This article explores the classic Speed Dating dataset, performing data cleaning, exploratory analysis of match rates, gender and age effects, correlation studies, and finally building a logistic regression model with SVMSMOTE oversampling to predict matchmaking success, achieving around 83% accuracy.
This article presents a comprehensive study of tree-based deep models for large-scale matching, introduces the theoretical framework of optimal tree models, proposes a Beam Search aware training algorithm (BSAT/OTM) to address training-test mismatch, and demonstrates significant recall improvements on Amazon Books and UserBehavior datasets.
An open-source Java Spring Boot project demonstrates license plate detection and recognition using OpenCV, supporting multiple plate colors, with SVM and ANN training, detailed architecture, feature list, installation guide, and visual processing steps, offering a beginner-friendly tutorial for image recognition enthusiasts.
This article presents CSCNN, a category‑specific convolutional neural network that integrates visual priors into click‑through‑rate (CTR) models for JD.com’s e‑commerce advertising, detailing its motivation, architecture, engineering optimizations, offline and online training strategies, and empirical performance gains on both public and industrial datasets.
In this interview, Peng Changping, JD's recommendation advertising algorithm lead, explains how large‑scale machine learning and deep learning power JD's e‑commerce recommendation system, detailing its impact on revenue, key technical components, challenges, and future directions.
This article details the end‑to‑end deep‑learning workflow for click‑through‑rate (CTR) prediction in 58.com’s commercial ranking system, covering system architecture, feature engineering, sample construction, model evolution from Wide&Deep to DIN/DIEN, and engineering optimizations that together yielded significant CPM and CVR improvements.
This article presents twenty classic NumPy problems covering array lookup, modification, conversion, sampling, slicing, string operations, rounding, reshaping, linear algebra, and more, each accompanied by concise Python code examples and visual illustrations to help you master advanced data manipulation techniques.
This article explains the fundamental differences between computational advertising and recommendation systems, outlines the distinct problems each field addresses, and surveys the most widely used advertising models—including traditional machine‑learning approaches, deep‑learning architectures, and hybrid solutions—providing practical insights for engineers in both domains.
Tencent engineers broke the world record for training ImageNet with 128 V100 GPUs in just 2 minutes 31 seconds, detailing a suite of optimizations—including a new Light distributed training framework, single‑machine speed boosts, multi‑machine communication enhancements, and advanced batch convergence techniques—that together dramatically cut training time while maintaining high accuracy.
This article introduces Alibaba Cloud’s free PAI‑DSW cloud IDE for AI development, explains the evolution of machine learning, guides users through creating notebooks, running Python code, and demonstrates a complete speed‑dating dataset analysis and predictive modeling pipeline using logistic regression and data‑balancing techniques.
This article explains how Fliggy's travel recommendation platform tackles recall challenges such as cold‑start users, sparse behavior, itinerary‑specific needs, and periodic repurchase by applying user‑attribute models, graph embeddings, dual‑tower architectures, session‑based methods, and statistical repurchase forecasting to improve candidate selection and overall recommendation performance.
JD Digits' intelligent anti‑fraud platform leverages machine learning, big‑data processing, graph neural networks and small‑sample knowledge‑graph algorithms to provide millisecond‑level, real‑time protection across 600+ scenarios, while also offering AI‑powered solutions to banks and publishing research at top conferences.
This article shares Weibo’s experience in building and evolving its recommendation algorithms, covering the recommendation scenario, machine learning workflow, feature engineering, model upgrades, large‑scale challenges, deployment via the Weiflow platform, and the capabilities of its machine‑learning infrastructure.
The winning KDD Cup 2020 debiasing solution builds a heterogeneous item‑to‑item graph with click‑co‑occurrence and multimodal similarity edges, uses multi‑hop random walks to generate unbiased candidate samples, trains LightGBM with a popularity‑weighted loss, and aggregates scores to lift low‑popularity items, thereby eliminating selection and popularity bias and achieving first place among 1,895 teams.
This article presents a detailed four‑year study plan for aspiring artificial‑intelligence professionals, curated by Stanford graduate Mihail Eric, listing foundational, intermediate and advanced courses—including CS 106B, CS 107, CS 221, CS 229, CS 231N and more—along with practical advice for projects, research and internships.
This article explains the node2vec algorithm, its mathematical foundations, biased random‑walk sampling strategy with parameters p and q, implementation details using the Alias method, and demonstrates its superior performance on node classification and visualization tasks compared with DeepWalk and LINE.
The article introduces COLD, a computing‑power‑aware online and lightweight deep pre‑ranking system for Alibaba's targeted ads, detailing its evolution from static CTR models to vector‑inner‑product models, its flexible network architecture with feature‑selection via SE blocks, engineering optimizations such as parallelism, column‑wise computation, Float16 and MPS, and demonstrates superior offline and online performance through extensive experiments.
This article presents a hand‑drawn, illustrated walkthrough of the entire machine‑learning pipeline—from dataset definition, exploratory data analysis, preprocessing, and data splitting to model building, algorithm selection, hyper‑parameter tuning, feature selection, and evaluation for both classification and regression tasks.
GoodFuture's AI team secured the UbiComp2020 championship by using a novel sensor‑driven handwriting recognition approach that bypasses traditional OCR, showcasing China's growing strength in educational AI and advanced deep‑learning models.
This article presents a set of charts outlining various learning paths and technologies for aspiring machine‑learning engineers, offering guidance on what to study and why, while also providing community resources for deeper discussion and support.
This article introduces chatbots, explains rule‑based and self‑learning types, describes how AI‑driven bots work, and provides a complete Python tutorial using the ChatterBot library—including environment setup, installation, and full source code for a functional chatbot.
This article presents the design, implementation, and performance of 58.com’s dialogue robot for local life services, detailing its business background, overall AI architecture, core QABot and TaskBot capabilities, intelligent lead generation, opportunity mining, and future research directions.
This article explores the career bottlenecks faced by programmers over 30, presents three practical transition paths—including project management, tech evangelism, and AI development—and provides a detailed, no‑code tutorial for building a facial‑recognition model with Huawei Cloud ModelArts.
This article explains the fundamentals of user profiling in the big data era, covering tag types, data architecture, development modules, a step‑by‑step implementation process, a practical e‑commerce case study, table design strategies, and both quantitative and qualitative profiling methods.
This article presents 58.com’s intelligent voice quality inspection system, detailing its overall architecture, speech separation, speaker role identification, NLP‑based tagging, model choices such as VGG, BERT, ALBERT and SPTM, and real‑world call‑center use cases that improve efficiency and reduce risk.
This article details the technical architecture behind WeChat's 'Kan Kan' content understanding platform, covering text and multimedia analysis, tag extraction, entity recognition, knowledge graph construction, and how these components enhance recommendation recall, ranking, and user engagement across the ecosystem.
The article details the Tencent Advertising Algorithm Competition semifinal, presenting the Embedding team's pure Transformer model with joint ID word embeddings, data preprocessing, training hyperparameters, model architecture, and practical tips for improving age/gender prediction from click sequences.
This article explains the principles and implementation of face detection based on statistical methods, detailing Haar feature types, integral image computation, feature normalization, cascade classifiers, and provides step‑by‑step OpenCV code examples for static images, eye detection, and real‑time webcam detection.
This guide walks you through setting up Anaconda, preparing UI component images, training a SimpleVGGNet model on macOS, and deploying the model as a Docker‑based service that can be queried via an HTTP API.
Meituan’s AI‑driven search combines the Augur online inference framework—offering stateless, distributed feature operators, transformers, and a DSL for rapid, high‑throughput model scoring—with the Poker platform for model training, versioning, and experimentation, together accelerating iteration, improving performance, and enabling advanced model‑as‑feature ensembles.
The Qunar Technology Carnival features in‑depth interviews with experts Wang Mingyou, He Yang, and Jia Ziyan who share practical experiences on search ranking improvements, AIOps‑driven fault localization, and data‑driven revenue management, highlighting challenges, solutions, and future directions in AI‑powered systems.
The Meituan Search team boosted ranking relevance by training a domain‑specific BERT, applying data augmentation, brand‑sample optimization, knowledge‑graph fusion, multi‑task and pairwise fine‑tuning, joint end‑to‑end training with LambdaLoss ranking models, and compressing the model for low‑latency inference, delivering up to +925 BP offline accuracy gains and measurable CTR and NDCG improvements in production.
