This article presents a comprehensive overview of Taobao Live's deep ranking system evolution, detailing the DBMTL multi‑task learning framework, the two‑tower DMR matching‑ranking architecture, and the RUI Ranking refer‑item model, together with their offline formulas, online deployment scenarios, and measured performance gains across click‑through, watch‑time, and conversion metrics.
This article details the engineering challenges of serving a large BERT model in real‑time for 360 Search and describes a series of optimizations—including TensorRT‑based kernel fusion, model quantization, knowledge distillation, multi‑stream execution, caching, and dynamic sequence handling—that together achieve low latency, high throughput, and stable deployment on GPU clusters.
This article describes how 58.com’s rental platform leverages deep learning models such as Wide&Deep, DeepFM, DCN, DIN, and DIEN to improve search ranking, detailing data pipelines, feature engineering, model iteration, multi‑task training, prediction optimizations, and resulting online performance gains.
deep learningfeature engineeringmodel optimization
This article presents the end‑to‑end design, feature engineering, model evolution (Wide&Deep, DeepFM, DCN, DIN, DIEN), multi‑task training, and deployment optimizations that 58.com applied to improve search ranking for its rental business, demonstrating significant gains in click‑through and conversion rates.
Dewu Photo Search evolved from a limited Aliyun‑based prototype to a self‑developed pipeline using EfficientNet detection and 128‑dim embeddings, boosting top‑1 shoe accuracy over 100 % and overall precision by up to 41 %, while reducing latency and improving scalability despite remaining stability challenges.
This article presents a comprehensive overview of Ctrip's multilingual machine translation platform, detailing demand analysis, system architecture, data pipeline, algorithmic innovations such as task‑space fusion and term‑translation interventions, as well as extensive performance optimizations for low‑resource languages.
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 analyses the drawbacks of using fully‑connected layers in convolutional neural networks for image tasks, proposes fully‑convolutional alternatives with 1×1 convolutions and strategic max‑pooling, provides TensorFlow code examples, compares model sizes and performance, and discusses deployment considerations for variable‑size inputs.
This article describes the design, model selection, experimental evaluation, and engineering optimization of an AI‑driven post‑sale question‑answering system for hotel services, covering FAQ construction, intent detection, deep‑learning matching models such as DSSM, ESIM, BERT, and their performance and latency trade‑offs.
In 2020, the AR “scan‑for‑fortune” feature achieved a full AutoML rollout on the xNN‑Cloud platform, automating network architecture design and the entire model development pipeline, which cut Android inference time by over 50%, iOS by 30%, reduced model size, and boosted accuracy by 1.6% while handling billions of in‑client inferences.
The paper presents a lightweight, anchor‑free CenterNet‑based object‑ness detector for WeChat’s Scan feature, built on a ShuffleNetV2 backbone with enlarged 5×5 depth‑wise convolutions, a streamlined detection head, and a Pyramid Interpolation Module, then quantized, ONNX‑converted and NCNN‑deployed to achieve a 436 KB model running in ~15 ms per frame on an iPhone 8 CPU.
The study enhances UI2CODE’s ability to detect tiny UI components by augmenting training data with copied small objects, upgrading the detector from Faster RCNN to FPN and Cascade FPN, and refining box positions with smoothing and projection, achieving superior small‑object mAP/mAR and enabling broader UI parsing applications.
This article explains how the 58 AI platform leverages NVIDIA TensorRT to accelerate deep‑learning inference on GPUs, describes three integration approaches, details the TF‑TRT implementation and Kubernetes deployment, and presents performance gains for ResNet‑50 and OCR models.
This article explains how to use NVIDIA TensorRT to accelerate TensorFlow model inference by describing TensorRT architecture, optimization techniques such as layer fusion and precision calibration, detailing the conversion of frozen_graph and saved_model formats, presenting experimental setup and performance comparisons, and summarizing the achieved speed‑up.
Alibaba’s MNN (Mobile Neural Network) engine, now open‑sourced on GitHub, showcases how a lightweight, end‑side deep‑learning inference framework tackles fragmentation, optimizes model conversion, scheduling, and execution across diverse devices, delivering significant performance gains for mobile and IoT AI applications.
Gaode Maps applied Learning to Rank to optimize search suggestions, moving from rule-based to gradient boosted rank model, addressing sample construction and feature sparsity via session-based labeling and loss adjustment, achieving a seven‑point MRR gain and higher coverage, and paving the way for personalization and deep learning.
SmileAR, iQIYI’s self‑developed mobile AR platform powered by TensorFlow Lite, delivers real‑time face, body and gesture recognition across iQIYI’s apps through MobileNet‑based models, quantization‑aware training, multi‑task learning and encrypted SDKs, achieving fast, lightweight, cross‑platform AR experiences for millions of users.
This article details how 58.com improved the efficiency of its search list ranking by moving from simple time‑based ordering to a comprehensive ranking framework that incorporates feedback strategies, basic machine‑learning models, feature upgrades, and advanced model upgrades, achieving significant gains in click‑through, conversion, and revenue across multiple business lines.
The article presents a comprehensive overview of DianRong Net's in‑house machine‑learning platform built on Spark, its workflow, pain points it addresses, risk‑control case studies using graph mining, and practical tips for improving model performance through data, algorithms, hyper‑parameter tuning and ensemble methods.
Advancing a machine‑learning project requires focusing first on the core business problem, then designing comprehensive features, ensuring high‑quality data, and finally selecting an appropriate model, because business impact drives success while features and data set the performance ceiling and model choice balances accuracy with interpretability.
This article breaks down the four essential elements of a model project—business, features, data, and model—explains their priority order, and provides practical guidance on aligning technical choices with core business goals to achieve efficient, high‑impact outcomes.
AI project managementbusiness alignmentdata quality