Apache
/
lucene
mirror of https://github.com/apache/lucene.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

SOLR-15193: Improve maxDocFreq docs

This commit is contained in:
Joel Bernstein 2021-03-07 20:31:07 -05:00
parent 606cea94d7
commit 140c37eb0f
1 changed files with 17 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

27
solr/solr-ref-guide/src/graph.adoc
View File

@ -176,7 +176,7 @@ The ancestor links will only be tracked when the trackTraversal flag is turned o
Link analysis is often performed to determine *node centrality*. When analyzing for centrality the Link analysis is often performed to determine *node centrality*. When analyzing for centrality the
goal is to assign a weight to each node based on how connected it is in the subgraph. goal is to assign a weight to each node based on how connected it is in the subgraph.
There are different types of node centrality. Graph expressions very efficiently calculates There are different types of node centrality. Graph expressions very efficiently calculates
*inbound degree centrality* (indegree). *inbound degree centrality* (in-degree).
Inbound degree centrality is calculated by counting the number of inbound Inbound degree centrality is calculated by counting the number of inbound
links to each node. For simplicity this document will sometimes refer links to each node. For simplicity this document will sometimes refer
@ -274,17 +274,24 @@ image::images/math-expressions/graph2.png[]
If we compute the dot product between the butter column and the other product columns you will find that the dot product equals the inbound degree in each case. If we compute the dot product between the butter column and the other product columns you will find that the dot product equals the inbound degree in each case.
This tells us that a nearest neighbor search, using a maximum inner product similarity, would select the column with the highest inbound degree. This tells us that a nearest neighbor search, using a maximum inner product similarity, would select the column with the highest inbound degree.
=== Limiting Basket Size === Limiting Basket Out-Degree
The recommendation can be improved if we chose baskets that contain fewer items. The recommendation can be made stronger by limiting the *out-degree* of the baskets. The out-degree is the
This is because baskets with a smaller number of products carry more information about the number of outbound links of a node in a graph. In the shopping basket example the outbound links
relationship between the products in the basket. from the baskets link to products. So limiting the out-degree will limit the size of the baskets.
The `maxDocFreq` parameter can be used to limit the "walk" to only include baskets that appear in the index a certain Why does limiting the size of the shopping baskets make a stronger recommendation? To answer this question it helps
number of times. Since each occurrence of a basket ID in the index is a product, limiting the document frequency of the to think about each shopping basket as *voting* for products that go with *butter*. In an election with two candidates
basket ID will limit the size of the basket. The `maxDocFreq` param is applied per shard. If there is a single if you were to vote for both candidates the votes would cancel each other out and have no effect.
shard or documents are co-located by basket ID then the `maxDocFreq` will be an exact count. But if you vote for only one candidate your vote will affect the outcome. The same principal holds true
Otherwise it will return baskets with a max size of numShards*maxDocFreq. for recommendations. As a basket votes for more products it dilutes the strength of its recommendation for any
one product. A basket with just butter and one other item more strongly recommends that item.
The `maxDocFreq` parameter can be used to limit the graph "walk" to only include baskets that appear in
the index a certain number of times. Since each occurrence of a basket ID in the index is a link to a product,
limiting the document frequency of the basket ID will limit the out-degree of the basket. The `maxDocFreq` param is
applied per shard. If there is a single shard or documents are co-located by basket ID then the `maxDocFreq` will
be an exact count. Otherwise, it will return baskets with a max size of numShards * maxDocFreq.
The example below shows the `maxDocFreq` parameter applied to the `nodes` expression. The example below shows the `maxDocFreq` parameter applied to the `nodes` expression.