# 向量資料庫 ## 目錄 [TOC] ## 甚麼是向量? ![image](https://hackmd.io/_uploads/HJZ-rqGAA.png) ![image](https://hackmd.io/_uploads/S1wuBoz00.png) ## 甚麼是Embedding? ![image](https://hackmd.io/_uploads/B1Mprif00.png) ![image](https://hackmd.io/_uploads/rJ11UiM0C.png) ![image](https://hackmd.io/_uploads/HJ8xLozA0.png) ![image](https://hackmd.io/_uploads/B19vIszCA.png) ![image](https://hackmd.io/_uploads/HJtY8of0A.png) ![image](https://hackmd.io/_uploads/HyZcLiMCA.png) ## 向量資料庫跟傳統資料庫區別 ![image](https://hackmd.io/_uploads/H1Ij16zAA.png) ![image](https://hackmd.io/_uploads/r1arPof0C.png) ![image](https://hackmd.io/_uploads/HkywwjMRA.png) ![image](https://hackmd.io/_uploads/HkcvwjfRC.png) ![image](https://hackmd.io/_uploads/ByDcwiz0A.png) ## 當前常用向量資料庫 ![image](https://hackmd.io/_uploads/SJnag6GRC.png) ## 實際應用 ![image](https://hackmd.io/_uploads/BJQ9WafRR.png) ### 實作範例 #### 實作一 ##### [faiss](https://colab.research.google.com/drive/1_4hhQgW3bWetfGtyHRh8_7gwezdPEV6x) #### 實作二 ##### [chroma db](https://github.com/princend/face_recognition_opencv/tree/vector-db) ## BONUS - HNSW 演算法 ![image](https://hackmd.io/_uploads/Sy0hYCMR0.png) ## Reference ### 向量 - [高中向量介紹影片-均一教育平台](https://www.youtube.com/watch?v=dhwlNUDHC-4&list=PLp2Y5q36tB-Psi4jKfg-S73tbBvfOZWxc&ab_channel=均一教育平台JunyiAcademy) ### Embedding - [word embedding](https://www.youtube.com/watch?v=X7PH3NuYW0Q&ab_channel=Hung-yiLee) ### 向量資料庫 - [基礎介紹](https://aws.amazon.com/tw/what-is/vector-databases/) - [基礎介紹2](https://blog.darkthread.net/blog/vector-db-concept/) - [基礎介紹3](https://tenten.co/learning/vector-database/) - [基礎介紹-YT影片](https://www.youtube.com/watch?v=cv_x2hvbZrQ&list=PLuufbYGcg3p69q_OJoC6USZNf6dK4tXcf&ab_channel=ZOMI酱) - [基礎介紹-YT影片的講義](https://github.com/chenzomi12/AISystem/tree/784404f22ab6d0b949e4583cab648f3c184ac8f6/06Foundation/05Dataset) - [生成式AI(11) — 本地的向量資料庫Chroma + OpenAI](https://medium.com/@King610160/生成式ai-11-本地的向量資料庫chroma-openai-7882e54c1fb2) - [pinecone範例](https://colab.research.google.com/github/pinecone-io/examples/blob/master/docs/semantic-search.ipynb#scrollTo=dJbjE-iq_yMr) - [chroma db範例](https://github.com/billlin0904/AudioEmbeddingQuery) ### ANN演算法 - [ANN-algo](https://wanger-sjtu.github.io/ANN-algo/) - [HNSW YT影片1](https://www.youtube.com/watch?v=QvKMwLjdK-s&ab_channel=JamesBriggs) - [HNSW YT影片2](https://www.youtube.com/watch?v=77QH0Y2PYKg&ab_channel=DataMListic) - [HNSW YT影片3](https://www.youtube.com/watch?v=_pewMGesffk&ab_channel=MahamMukaram) - [connected graph](https://web.ntnu.edu.tw/~algo/ConnectedGraph.html) - [NSW與HNSW差異](https://chatgpt.com/share/66f55b2e-edec-8003-a001-467416822fbf) - [KD-Tree](https://blog.yucheng.me/post/kd-tree/) - [BM25](https://myapollo.com.tw/blog/bm25-explanation/) - [B-Tree](https://zh.wikipedia.org/zh-tw/B树)