milvus-io
/
milvus
mirror of https://github.com/milvus-io/milvus.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Refine words in proxy reduce md file (#12606) 

Signed-off-by: yanliang567 <yanliang.qiao@zilliz.com>
pull/12702/head
yanliang567 2021-12-03 15:19:43 +08:00 committed by GitHub
parent d1be80ff83
commit cb70c30b96
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 4 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
docs/developer_guides/proxy-reduce.md
View File

@ -1,16 +1,16 @@
# How Proxy Reduce the Multiple Search Results?
# How Proxy Reduces the Multiple Search Results?
## Search Results Organized format
For a specified Search request, with nq = 2 and topk = 10, Proxy received 4 search results from query nodes.
Each Search Result is nq \* topk two-dimensional structure，as the illustration below. The result indicates that the user has input NQ vectors and wants to obtain the TOPK most similar vectors to these NQ vectors respectively.
Each Search Result is nq * topk two-dimensional structure，as the illustration below. The result indicates that the user has input nq vectors and wants to obtain the topk most similar vectors to these nq vectors respectively.
![search_result_format](./figs/nq_topk_search_results.png)
For each query, the top k hit results are in descending order of score. The larger the score, the more similar the hit result is to the vector to be queried. The hit results of different queries are independent of each other.
For each query, the topk hit results are in descending order of score. The larger the score, the more similar the hit result is to the vector to be queried. The hit results of different queries are independent of each other.
Therefore, we will only discuss how the proxy merges the results for one query result. For NQ query results, we can loop through NQ or process them in parallel.
Therefore, we will only discuss how the proxy merges the results for one query result. For nq query results, we can loop through nq or process them in parallel.
So the problem degenerates to how to get the maximum number of 10 (TOPK) results from these four sorted arrays. As shown in the figure below:
![final_result](./figs/reduce_results.png)