In a controlled cold‑start test of a simple hello‑world script on the same machine, Bun 1.4 launches in 4.8 ms, outpacing Deno 2.9.3 (14.9 ms) and Node.js 26.5.0 (22.0 ms), a speed advantage of roughly 3‑4×, though the results apply only to this limited scenario.
This article explains the serverless execution model, compares major providers such as AWS Lambda, Vercel Functions, and Cloudflare Workers, and offers practical code examples and cold‑start mitigation techniques for building and operating serverless applications.
Building a Retrieval‑Augmented Generation (RAG) system from scratch without existing QA pairs requires a systematic cold‑start approach that creates synthetic QA data, establishes baseline metrics, iteratively improves via expert labeling and real user feedback, and ensures document quality for reliable evaluation.
After building a fast Rust microservice, the team discovered Kubernetes pods took 8‑10 seconds to start due to Alpine‑based images; switching to minimal Distroless containers and static linking reduced the image size from 40 MB to 6.7 MB, cut cold‑start time to ~1.2 seconds, lowered memory usage, and improved security.
This article details a systematic four‑phase optimization of React Native page startup—preloading resources, advancing initialization, reusing and pre‑creating containers—that shrank cold start time by 55 % (1.78 s → 0.80 s) and hot start time by 70 % (1.10 s → 0.33 s), while exposing the new pitfalls introduced by container reuse.
This article explains how advertising recommendation is evolving from traditional feature‑engineered models to LLM‑driven pipelines, detailing data‑infrastructure challenges, semantic upgrades with multimodal embeddings, case studies in short‑video ads, user cold‑start prompt engineering, and future directions for generative recommendation systems.
A detailed benchmark of a reverse‑shell project shows that Go outperforms Node.js in cold‑start latency, memory consumption, and binary size, while Node.js narrows the gap on warm‑path latency, highlighting the trade‑offs developers must weigh when deciding to rewrite.
The paper introduces SSR, a novel multimodal recommendation framework that leverages graph Fourier transforms, energy‑balanced frequency bands, structured regularization, and low‑rank tensor decomposition to replace black‑box fusion with explainable, adaptive reasoning, achieving state‑of‑the‑art results on Amazon datasets and strong cold‑start performance.
This article explains a SIGIR 2025 paper that tackles cold‑start click‑through‑rate prediction in JD's ad system by using a Graph Isomorphism Network‑based cohort modeling framework, detailing its three‑module architecture, extensive experiments on public and industrial datasets, and a live deployment that achieved a 2.13% CTR lift.
The article analyzes the FilterLLM approach, which augments a frozen LLM with billions of learnable user tokens to predict a full‑user interaction probability distribution in a single forward pass, dramatically speeding up cold‑start recommendation while preserving recommendation quality across multiple benchmarks.
DeepSeek‑R1 demonstrates that large‑scale reinforcement learning, especially with the novel Group Relative Policy Optimization and a rule‑based reward scheme, can markedly boost reasoning in LLMs without heavy supervised fine‑tuning, while a brief cold‑start SFT phase, two‑stage alignment, and knowledge distillation further improve performance and efficiency, despite remaining challenges such as language mixing.
This article details how NetEase Cloud Music leverages multimodal content understanding—using audio models like MusicCLIP and Audio MAE and image‑text fusion via FLAVA—to improve recommendation performance for new content and new users, covering system architecture, cold‑start solutions, and future AI‑driven directions.
This article reviews the Kuaishou paper on adapting large language models for recommendation, detailing the LEARN framework's dual‑tower architecture, embedding generation, loss functions, and experimental results that address cold‑start and long‑tail challenges in modern recommender systems.
This article analyzes Kuaishou’s May 2024 paper on LLM‑driven recommendation, detailing its dual‑tower architecture, contrastive learning of user and item embeddings, and a CVR‑auxiliary task that together improve cold‑start handling and boost both offline and online AUC metrics.
This article surveys the impact of large language models on recommendation systems, outlines generative recommender architectures, discusses challenges of JD Union advertising, presents a semantic‑ID based solution with training and inference details, and reports offline and online experimental results.
This article surveys how large language models can reshape recommendation systems, describes the four-stage generative pipeline, details item representation techniques such as semantic IDs, presents a JD affiliate advertising use case with offline and online experiments, and outlines future optimization directions.
This article reviews the NoteLLM paper, which leverages Llama 2 to create richer text embeddings and automatically generate tags and categories for note recommendation, addressing cold‑start issues through a multitask prompt design, generative‑contrastive learning, and collaborative supervised fine‑tuning, and demonstrates strong offline and online gains.
This article presents Baidu's comprehensive approach to cold‑starting new content in its large‑scale feed recommendation system, covering the definition and challenges of content cold start, algorithmic practices, ID feature optimizations, traffic control mechanisms, experimental design, and key Q&A insights.
This article details Alipay's advertising algorithm upgrades—including sample‑enhanced cold‑start mitigation, cross‑scene and user‑segmented knowledge transfer, and real‑time feature and online‑learning optimizations—that collectively lifted CTR, CPM, and overall business revenue.
The advertising engine uses a five‑stage funnel—retrieval, recall, coarse ranking, fine ranking, and re‑ranking—each optimized with specialized indexes, multi‑channel recall, multi‑objective twin‑tower models, deep CTR/CVR predictors, and cold‑start paths, delivering up to 33 % spend growth, 6 % eCPM lift and lower latency while maintaining diversity.
The article explains why first‑frame latency matters for video apps and presents a comprehensive set of optimization methods—including pre‑fetching URLs, intelligent preloading, prerendering, and scenario‑specific tweaks for cold start, short‑video scrolling, page navigation, and long‑video playback—to dramatically reduce start‑up time and improve user experience.
Xiaohongshu’s recommendation system uses large‑scale multimodal embeddings, dual‑tower and graph models, and diversity techniques like DPP and SSD to quickly surface high‑quality user‑generated content, enabling ordinary users to gain visibility while balancing personalization, exploration, and efficient LLM‑augmented pipelines.
This article presents NetEase Cloud Music's multimodal cold-start recommendation approach, detailing the problem's significance, feature extraction using CLIP, I2I2U indirect modeling, U2I DSSM direct modeling with contrastive learning and interest‑boundary mechanisms, deployment pipeline, evaluation results, and future optimization directions.
The article details how Autohome tackled the cold‑start problem for newly generated content by redesigning the recommendation pipeline, introducing multi‑path recall, refining ranking and re‑ranking formulas, and applying strategic controls, resulting in a rise of cold‑start success rate from 27% to over 99% and a CTR increase from 5% to 14%.
At CNCC 2023, leading researchers and industry experts convened to examine how large language models can transform recommendation systems, outlining four core challenges—model integration, fluency versus intelligence, hallucination versus deception, and user understanding—while highlighting opportunities such as multimodal content, cold‑start solutions, zero‑shot ranking, instruction‑driven algorithms, and responsible, interactive recommendation pipelines.
This article investigates how large language models can be applied to recommendation tasks, presenting two usage strategies, experimental evaluations on multiple datasets, comparisons with traditional baselines, and analyses of prompting methods, cost, and cold‑start performance.
This article presents Kuaishou's approach to solving the content cold-start problem by analyzing its impact on video growth, detailing the challenges of sparse and biased training data, and describing a suite of graph‑neural‑network, I2U/U2I, TDM, and debiasing techniques that improve early video exposure and long‑term ecosystem health.
CC‑GNN addresses three drawbacks of existing graph‑neural retrieval for e‑commerce by adding content phrase nodes, scalable meta‑path message passing, and difficulty‑aware noisy contrastive learning with counterfactual augmentation, achieving up to 16 % recall improvement and notably larger gains on long‑tail queries and cold‑start items.
The DataFunSummit 2023 online summit (August 26‑27) will explore eight recommendation‑system topics—including core and engineering architecture, model training/inference, large models, graphs, cold start, and multi‑task scenarios—featuring Xiaohongshu leaders who will present on graph‑based business architecture, integrated training‑inference pipelines, and user/content cold‑start strategies.
This article examines a real‑world case where a serverless instance’s automatic scaling was instantly overwhelmed, causing high CPU usage, frequent Full GC and JVM crashes, and then demonstrates how Sentinel’s system‑level rules can mitigate the overload and improve cold‑start performance.
The article analyzes key challenges in large‑language‑model pipelines—including the necessity of cold‑start pretraining, the pitfalls of reward‑model hacking, efficiency‑effectiveness trade‑offs, evaluation difficulties, and downstream fine‑tuning limits—offering practical insights for more reliable LLM development.
This article reviews how recent large language models (LLMs) are reshaping recommendation systems, covering their emergence, in‑context learning, prompt‑based strategies, three main LLM‑driven recommendation paradigms, key research papers, experimental results, and future research directions.
Social4Rec introduces a social interest‑enhanced video recommendation framework that tackles user cold‑start by extracting coarse‑ and fine‑grained social interests via a self‑organizing neural network and meta‑path neighborhood aggregation, integrating these embeddings with a YouTube DNN model to improve CTR and AUC.
This article proposes a graph‑based node representation method that combines static attribute graphs and dynamic interaction graphs with multi‑level attention to alleviate user and item cold‑start problems in recommendation systems, achieving notable AUC improvements on sparsified MovieLens datasets.
This article presents EGES, a graph‑embedding model that incorporates side information to construct a directed user graph, apply biased random‑walk sampling, and train weighted Skip‑Gram embeddings, thereby improving large‑scale user acquisition and addressing cold‑start challenges in recommendation systems.
This article explains what recommendation systems are, their core tasks, evaluation metrics, popular algorithms such as collaborative filtering and latent factor models, how to handle cold‑start and contextual challenges, the role of social networks, and typical system architecture, providing a comprehensive overview for beginners and practitioners.
This interview compiles expert opinions on the end‑to‑end recommendation system pipeline—including architecture, data collection, user profiling, content structuring, feature engineering, recall strategies, ranking algorithms, multi‑objective optimization, multi‑modal fusion, re‑ranking, cold‑start solutions, evaluation metrics and real‑world applications—highlighting the technical challenges and practical solutions.
This interview with three recommendation‑system experts explores the technical architecture, data sources, feature engineering, recall and ranking strategies, evaluation metrics, cold‑start solutions, and practical difficulties, offering actionable insights to avoid common pitfalls in real‑world recommender deployments.
This article explains how Alipay measures and improves Android cold‑start performance through video‑frame analysis, ActivityTaskManager, custom instrumentation, home‑page snapshot techniques, temperature control, patch‑APK data collection, AOP‑based diagnostics, and an integrated APM platform.
Alipay’s client engineering team details a comprehensive approach to monitoring, measuring, and improving time‑consuming user experiences—especially cold‑start—by employing video frame analysis, ActivityTaskManager, extensive instrumentation, home‑page snapshot techniques, temperature control, patch‑APK injection, AOP‑based diagnostics, and the Spider SDK within a robust APM platform.
This article details the evolution of the ‘逛逛’ content community’s recommendation system, comparing the legacy rule‑based Hive workflow with a new algorithm‑driven architecture that leverages Elasticsearch, Redis, multi‑stage recall, coarse‑ and fine‑ranking, re‑ranking, exposure filtering, cold‑start handling, performance tuning, and future plans for vector‑based recall and platformization.
The article details Guangguang’s shift from a rule‑based, Hive‑driven recommendation pipeline to an algorithmic service that leverages Elasticsearch and Redis for multi‑source recall, coarse and fine model ranking, exposure filtering, cold‑start handling, latency optimizations, reliability monitoring, and future vector‑based enhancements.
This paper introduces MAIL, a model‑agnostic dual‑tower framework that uses a zero‑shot learning tower to generate virtual user behaviors for new users and an embedding‑based ranking tower, achieving 13‑15% CTR lift in large‑scale live‑stream recommendation at NetEase Cloud Music.
This article explains why cold‑start is a critical challenge for recommender systems, outlines four practical strategies—generalization, fast data collection, transfer learning, and few‑shot learning—and then details the DropoutNet model, its end‑to‑end training, loss functions, negative‑sampling techniques, and open‑source implementation.
The article examines cold‑start challenges for rich‑media recommendations, outlines detection via calibration and lifecycle monitoring, and proposes two remedies—the multi‑stage “rise channel” for promoting fresh content and cross‑modal understanding using CLIP, CB2CF and dual‑tower models—demonstrating NetEase Cloud Music’s 25% distribution boost, over 20% CTR rise, and 40% review‑work reduction.
The article presents NetEase Cloud Music’s tag‑based cold‑start recommendation approach, which generates song tags, computes similarity, and recommends new, unevaluated tracks to users, addressing data scarcity and quality challenges while improving exposure coverage and balancing user experience, establishing a vital foundation for a healthy music ecosystem.
The paper tackles Taobao Live’s cold‑start problem for new users by introducing a multi‑channel lifelong product‑sequence network that enriches purchase histories with side information, extracts relevance‑focused subsequences across five channels, and integrates them via target‑attention DIN, achieving substantial offline and online performance gains, especially for low‑activity users.
This article explains the fundamentals of Android app startup, categorizes cold, hot, and warm launches, identifies performance bottlenecks using tools like ADB, Systrace, and Traceview, and presents concrete optimization strategies for Application creation and splash-screen rendering that reduce launch time by up to 60 percent.
This article details NetEase Cloud Music's technical approach to building implicit user relationship chains—using SimHash, Item2Vec, and MetaPath2Vec embeddings, large‑scale vector search, and a unified service architecture—to address cold‑start challenges across multiple business scenarios.
This article details the background, analysis, and step‑by‑step solutions for reducing package size, improving cold‑start time, and establishing a comprehensive mobile performance testing framework for both Android and iOS versions of the Xueersi Online School app.
Alinode PGO leverages profile‑guided optimization to generate a cache of hot startup code for Node.js functions, cutting cold‑start latency by up to 61% (150% speed‑up) and allowing further reductions by launching solely from the PGO cache, as demonstrated with real‑world benchmarks.
This guide explains why Spring Boot applications suffer 30‑second cold‑start delays on serverless platforms, breaks down each startup phase, and provides three practical optimization techniques—reserved instances, lazy initialization with JVM tweaks, and proper instance‑concurrency sizing—to dramatically improve response times.
This article, part of a series analyzing Serverless platforms for Spring Boot, explains how to diagnose and reduce cold‑start latency, covering tracing, reserved instances, lazy initialization, JVM options, and instance concurrency tuning, using the high‑traffic Mall demo application as a concrete example.
This article presents a comprehensive overview of Feizhu's information‑flow recommendation system, detailing its mixed‑material architecture, cold‑start recall and coarse‑ranking techniques, multi‑modal pre‑training and fine‑tuning, fine‑ranking with user‑state gating, and tiered traffic‑flow mechanisms for content delivery.
This guide walks through analyzing and optimizing cold‑start latency for a Spring Boot e‑commerce app on Alibaba Cloud Function Compute,{} covering tracing, reserved instances, lazy initialization, JVM tuning, and instance concurrency settings to improve performance.
This article explains the root causes of Java cold‑start latency, describes how static compilation with GraalVM removes JVM initialization and JIT overhead, and shows performance comparisons that demonstrate near‑instant startup for serverless Java applications.
This talk explains how 58.com tackles the cold‑start and interest‑divergence problems of its massive blue‑collar job recruitment platform by modeling the recommendation process as a reinforcement‑learning task, detailing the use of multi‑armed bandit, contextual bandit, and linear‑UCB algorithms, offline evaluation pipelines, online deployment, and observed performance gains.
HetMatch is a heterogeneous graph neural network for keyword recommendation in search advertising that tackles cold‑start and large‑scale challenges by hierarchically fusing node and subgraph features, denoising graph convolutions, applying self‑attention, twin matching, and multi‑view learning, delivering notable recall gains and online performance improvements for Alibaba’s advertising tools.
This talk explains how 58.com’s massive blue‑collar recruitment platform uses reinforcement‑learning techniques—including multi‑armed bandits, contextual MAB, and linear UCB—to address cold‑start and interest‑divergence challenges, describes the system architecture, offline evaluation, online deployment, and reports an 8% uplift in new‑user conversion.
iQIYI’s short‑video team solves the new‑video cold‑start problem by using a GAN that generates latent user features from video attributes and a discriminator to validate them, then matches these vectors to real users via cosine similarity, achieving double‑digit gains in exposure, CTR, and watch time.
This article presents TRAINOR, a travel‑intention‑aware out‑of‑town POI recommendation framework that tackles cold‑start, interest‑drift, and geographical gaps by jointly modeling hometown preferences with graph neural networks, neural topic models for travel intention, and matrix‑factorization‑based out‑of‑town preference transfer, and validates its superiority through extensive cross‑city experiments.
This article analyzes why Alipay’s cold start on our device is slower than on competing phones and presents a series of optimizations—including CPU boosting, Dex2oat strategy adjustments, GC frequency reduction, and I/O improvements—that collectively close the performance gap.
This article dissects the Android cold‑start sequence of the Alipay app, detailing each stage from the initial input event through process forking, activity initialization, splash screen rendering, and final frame display, while providing timing data and optimization insights.
This article proposes TRAINOR, a travel‑intention‑aware out‑of‑town POI recommendation framework that tackles cold‑start and interest‑drift challenges by integrating graph neural networks for hometown preference, neural topic models for generic travel intentions, personalized intention inference, geographic modeling, and a preference‑transfer MLP, validated on real cross‑city check‑in data with superior recall performance.
This article presents Alibaba Fliggy's personalized marketing platform for travel, detailing its multi‑scene architecture, user‑session modeling, graph‑based recommendation algorithms, cold‑start strategies, cross‑domain user mapping, and a hierarchical travel‑play tag system that together enable large‑scale, real‑time, thousand‑person‑one‑face marketing.
This article systematically presents the cold‑start performance problems of the Beike iOS app, defines testing standards, explains the essence of optimization, lists common pitfalls, and details a comprehensive set of practical solutions ranging from lifecycle‑aware task scheduling and minimal‑set launchers to dynamic‑library lazy loading, compile‑time I/O elimination, static‑initializer handling, dead‑code removal, and monitoring standards.
The article summarizes a Serverless roundtable at ServerlessDays China 2021, where experts from AWS, Alibaba Cloud, ByteDance, and Tencent Cloud discuss Serverless definitions, market adoption gaps, standardization needs, large‑scale implementations, cold‑start challenges, multi‑cloud strategies, and future directions for the technology.
This article explains the iOS app launch lifecycle—including cold and hot starts—details launch‑time optimizations before and after the main function, introduces the fundamentals of virtual memory and ASLR, and provides a practical guide to binary reordering using order files and Clang instrumentation.
This article analyzes Android's class loading and verification process, identifies optimization opportunities to speed up cold starts, compares industry approaches, and presents a semi‑automated analysis tool along with practical solutions for handling verification failures in mobile apps.
This article explores Knative’s traffic routing and autoscaling mechanisms, detailing how requests are initially directed through the Activator during cold‑start, how VirtualService configurations evolve, and how newer versions shift traffic handling to Kubernetes Service/Endpoint layers, improving performance and decoupling gateway logic.
This article presents a comprehensive study on C‑side user identity recognition for internet platforms, addressing cold‑start and sample‑scarcity challenges by comparing keyword matching, XGBoost, TextCNN, a fusion model, and an improved Wide & Deep architecture, showing that the latter achieves the highest F1 score of 80.67%.
This article reviews the MiNet model, which leverages cross‑domain information by modeling long‑term, source‑domain short‑term, and target‑domain short‑term user interests with hierarchical attention and an auxiliary task to improve CTR prediction and alleviate cold‑start issues.
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.
This article presents a Neighborhood‑based Transfer Learning approach to solve the Dual Cold‑Start Recommendation problem in news services by transferring app‑installation similarity knowledge and using category‑level preferences to recommend unseen articles to brand‑new users.
By establishing measurable standards, profiling the launch flow, and applying task‑governance techniques—delaying, shrinking and splitting blocking tasks—plus home‑page shortcuts like ad pre‑fetch, background view creation and SharedPreferences caching, Xianyu’s Android cold‑start on low‑end phones dropped from about ten seconds to under five, with the first screen in ~1.3 seconds and full launch in 4.5 seconds.
The article examines the evolution of cloud computing, highlighting 2009’s pivotal milestones and the 2019 rise of Serverless, then critically assesses Serverless’s touted advantages—elasticity, reduced operations, rapid deployment—against practical challenges such as cold‑start latency, vendor hype, and migration complexities.
This article presents Alibaba Fliggy's personalized marketing platform for travel, detailing its architecture, scenario and functional abstractions, user‑modeling pipelines, full‑stack traffic control, cold‑start techniques, cross‑domain mapping, heterogeneous graph modeling, and a hierarchical travel‑play tag system to achieve thousand‑person‑one‑face recommendation across daily and promotional scenes.
The article outlines Android’s cold‑start sequence, describing how the system’s zygote, system_server services, and Activity Manager Service allocate resources, fork a new process, create the Application and Activity objects, render the first UI frame, and replace the placeholder window, helping developers diagnose and optimize launch performance.
This article presents a comprehensive overview of graph‑based recommendation techniques, including collaborative filtering, graph embedding, side‑information enhanced embeddings, two‑tower DSSM models, and practical cold‑start strategies from Alibaba and Airbnb, followed by a mixed model and Q&A session.
This article explains the cold‑start problem in recommendation systems, outlines common industry solutions such as popular‑content, group‑representative, auxiliary‑information, bandit algorithms, and deep learning, and details how Huajiao Live applied these techniques to improve new‑user engagement and metrics.
The article examines Android cold‑start performance by comparing three process‑creation timestamps—Application <init>, Process.getStartElapsedRealTime, and /proc/self/stat starttime—detailing their origins, extraction methods, conversion formulas, and practical guidance on which to use for accurate startup measurement.
The paper proposes a Siamese‑based PredictionNet that combines EfficientB3 image and VGGish audio features with user metrics to predict a HotValue score for newly uploaded short videos, using a margin loss with view‑value‑aware pair selection, enabling tiered cold‑start exposure that boosts overall platform efficiency.
This article explains how Alibaba Cloud Function Compute can be used for AI model serving, walks through a three‑step deployment of a TensorFlow‑based Chinese poem generator, compares serverless with traditional ECS setups, and discusses cold‑start mitigation, cost optimization, and monitoring features.
The CBVRP (Content-Based Video Relevance Prediction) challenge, co‑hosted by Hulu and ACM MM 2019, showcased the shift from user‑based collaborative filtering to content‑driven recommendation, highlighted winning teams and their papers, and underscored the ongoing research importance of cold‑start video recommendation for streaming platforms.
Youzan’s personalized recommendation platform combines a four‑layer architecture—data, storage, service, and application—with multi‑dimensional real‑time, offline, and cold‑start recall algorithms, Wide&Deep ranking, HBase/Druid storage, and configurable scene strategies to boost user conversion, traffic monetization, and future scalability.
This article analyzes why iOS apps experience long cold‑start times, explains the pre‑main and main launch phases, shows how to measure each stage using DYLD statistics and timestamp code, and provides concrete optimization strategies to reduce startup latency.
This article examines why Node.js serverless functions often exceed the desired 100 ms cold‑start target, analyzes profiling data to pinpoint costly file I/O and compilation steps, and presents practical optimizations such as module bundling, V8 code‑caching, and custom require handling to dramatically reduce startup latency.
This article details a comprehensive performance evaluation of Tencent Cloud Serverless functions, covering stress‑test methodology, cold‑start analysis, optimization techniques such as instance retention and code reduction, and real‑user latency comparisons, ultimately demonstrating how to achieve production‑grade response times.
This article explains how the MaFengWo iOS app defined startup, measured key metrics such as launch duration and loss rate, and applied technical and user‑centered optimizations—including pre‑main and post‑main tweaks, fine‑grained interaction strategies, ad caching, and platform mechanisms—to cut launch time by half, lower loss rate by 30% and dramatically increase ad exposure.
This article explains the fundamentals of the Word2Vec SGNS model, details its hyper‑parameters and training tricks, and demonstrates how customized embeddings are built for rental‑listing and news‑article recommendation, covering data preparation, objective‑function redesign, evaluation, and deployment in both recall and ranking stages.
This article presents a comprehensive overview of applying machine learning to recommendation systems, covering fundamental challenges such as user cold‑start, precise interest modeling, collaborative filtering, and both offline and online evaluation methods, while illustrating concepts with numerous diagrams.
This article details Meituan's systematic approach to reducing iOS app cold‑start time, covering the definition of startup phases, identification of performance bottlenecks, phased launch, self‑registration of startup tasks, code slimming, profiling tools, and continuous monitoring to achieve smoother user experience.
This article details the Ctrip Hotel ranking team's feature‑engineering and model‑innovation approaches—including session features, cold‑start mitigation, discriminative re‑weighting, and ensemble methods—that secured Top‑5 placements in the 2018 ACM WSDM and RecSys recommendation system competitions.
At the AI Pioneer Conference, Wang Xiaobo, head of Alibaba’s Commercial Machine Intelligence and Youku’s algorithm teams, discussed large‑scale distributed learning, recommendation challenges such as cold‑start and video heterogeneity, AutoML innovations, multi‑modal search during promotions, and the future demand for specialists in few‑shot learning and domain adaptation.
An in‑depth look at Alibaba’s billion‑scale graph embedding framework—GES and EGES—reveals how side‑information‑enhanced embeddings address user long‑tail coverage and cold‑start challenges, improving recommendation diversity and discovery across massive e‑commerce datasets and enabling real‑time personalized ranking.
Meitu’s personalized recommendation platform for the Meipai app combines offline feature engineering, near‑real‑time streaming, and online serving to recall, estimate, and rank billions of short videos using multi‑modal content features, user profiling, online learning, cold‑start bandit strategies, and multi‑objective diversity optimization, delivering timely, diverse feeds across live, homepage, and video‑detail scenarios.
This article explains how to measure and analyze Android SDK startup performance by defining cold, warm, and lukewarm launches, describing a testing methodology that uses specific device, network, and process conditions, and providing ADB commands and shell scripts to capture total launch time.
This article explains the concepts of cold, warm, and lukewarm Android app launches, outlines a testing methodology for measuring SDK startup times using ADB commands, and discusses practical considerations, test scenarios, and common pitfalls encountered during performance testing.
This article explains the objectives of recommendation systems, outlines four recommendation approaches, dives into personalized recommendation architecture and core algorithms, and discusses practical challenges such as real‑time processing, cold‑start, diversity, content quality, and exploration‑exploitation trade‑offs.
This article explains the objectives of recommendation systems, compares popular recommendation approaches, details the components and algorithms of personalized recommendation pipelines, and discusses practical challenges such as real‑time processing, freshness, cold‑start, diversity, content quality, and surprise handling.
