0 views collected around this technical thread.
This article demonstrates how to construct an AI‑powered medical knowledge base for diabetes treatment by preprocessing literature, performing semantic chunking, generating BioBERT embeddings, storing them in a FAISS vector database, and using a RAG framework together with a knowledge graph to retrieve and generate accurate answers.
This article details the end‑to‑end deep‑learning pipeline used by 58.com for real‑estate recommendation, covering business background, a six‑layer architecture, vector‑based recall, various embedding and ranking models, multi‑task and multi‑scenario optimization techniques, and future directions for large‑model integration.
This article details Bilibili's transition from a purchased customer‑service platform to a self‑developed system, describing the background, architectural design, core modules such as intelligent QA, seat scheduling, workbench, permission management, the use of Faiss for vector search, and future explorations with large language models, highlighting the technical challenges and solutions across backend development and AI integration.
The article details the architecture and techniques of ZhiZhuan's low-result search module, explaining how it combines ElasticSearch hard matching and sBert semantic vector soft matching, along with sophisticated negative sample strategies, to improve recommendation coverage and user experience.
This article explains the challenges of large‑scale vector retrieval, compares Faiss index types such as brute‑force, graph‑based and product quantization, and details how product quantization works, its memory‑speed trade‑offs, hierarchical quantization, and practical hyper‑parameter tuning.
The gpdb‑faiss‑vector plugin embeds Facebook’s Faiss library into the Dolphin (Greenplum‑compatible) engine, exposing SQL functions for distributed, high‑performance approximate nearest‑neighbor vector retrieval with caching, parallel search, configurable indexes, and sub‑millisecond latency, enabling scalable recommendation and advertising workloads.
The article reviews how high‑dimensional vector representations in deep‑learning applications require efficient approximate nearest‑neighbor search, comparing K‑d trees, hierarchical k‑means trees, locality‑sensitive hashing, product quantization, and HNSW graphs, and discusses practical FAISS implementations and how algorithm choice depends on data size, recall, latency, and resources.
This tutorial explains what Faiss is, how to install it, construct various indexes such as IndexFlatL2, IndexIVFFlat, and IndexIVFPQ, and demonstrates code examples for building and querying vector similarity search pipelines while discussing speed‑accuracy trade‑offs.
On December 6, Kuaishou and Tsinghua University’s joint team secured first place in the NeurIPS'21 Billion‑Scale Approximate Nearest Neighbor Search Challenge by leveraging a FAISS‑optimized, memory‑efficient KST_ANN algorithm that achieved over 6% higher recall on multiple billion‑scale datasets, showcasing the practical impact of large‑scale vector retrieval in AI‑driven services.
This article presents the end‑to‑end practice of a second‑hand housing recommendation system at 58.com and Anjuke, covering business background, embedding‑based vector recall, multi‑objective ranking methods such as ESMM and MMOE, experimental results, and future development directions.
This article presents a comprehensive case study of applying deep learning techniques—including item and user embedding, cross‑domain keyword intent modeling, and multi‑interest representation—to improve the recall stage of personalized recommendation for a cultural‑artifact auction platform, addressing unique data sparsity and diversity challenges.
This article details how 58.com improved its home‑page recommendation system by introducing vectorized recall with Word2Vec, optimizing negative sampling, deploying FAISS for fast nearest‑neighbor search, and later adopting a dual‑tower deep learning model with user interest features, achieving higher click‑through and conversion rates.
The article describes the design, architecture, and key components of a vector search platform built on Faiss that supports full‑index construction, incremental and distributed indexing, online retrieval, city‑level search, and vector update/delete operations to meet large‑scale AI application needs.
The iQiyi team “都挺好” tackled the 2019 CCF Video Copyright Detection contest by extracting frame‑level features with SE‑ResNeXt, indexing them with Faiss, aligning temporally via a critical‑path method, and refining copy boundaries using SIFT re‑matching and a sliding‑window approach, ultimately achieving an F1 score of 0.9678 after three iterative stages of model selection, cascade detection, and feature fusion.
This article details the commercial strategy team's exploration of embedding technologies for a second‑hand car platform, covering mainstream embedding methods, their application in advertising recall and ranking pipelines, system architecture, model optimizations, evaluation results, and future directions.
This article introduces the background of vector representations in machine learning, explains the K‑Nearest Neighbors algorithm and its key parameters, reviews traditional tree‑based and modern high‑performance search solutions, and demonstrates how FAISS can achieve microsecond‑level KNN queries on large‑scale high‑dimensional data.
The article explains how to replace keyword matching with deep‑learning document embeddings in Elasticsearch by applying PCA dimensionality reduction, indexing vectors using Lucene’s KD‑tree structures via a custom plugin, and leveraging FAISS‑style nearest‑neighbour techniques to achieve fast, semantically aware similarity search.
The Xianyu platform combats video plagiarism by extracting key frames, converting them into 1024‑dimensional vectors, and using product quantization‑based high‑dimensional vector retrieval to achieve over 95% recall with ~100 ms latency and more than 1000 QPS, enabling scalable video, image, and product deduplication.