TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Vespa.ai 5x faster than Elasticsearch at nearest neighbor ranking

8 点作者 andreer超过 5 年前

1 comment

bratao超过 5 年前
This confirm my real-world usage. We had to do a &quot;Semantic Similarity search&quot; for some documents, where we represent a document as a vector using BERT, and had to look for documents close to a reference document.<p>First we did on top of an ANN library(Facebook FAISS). For simplicity we moved to the naive Vespa vector search and the performance was good enough( FOR 1Mi documents, FAISS did the search in 50ms, Vespa was around 600ms), and we moved to Vespa to reduce the number of tools to maintain.