mirror of
https://github.com/hwchase17/langchain.git
synced 2025-06-21 22:29:51 +00:00
parent
4556b81b1d
commit
5686fed40b
@ -413,7 +413,7 @@
|
||||
"Yellowbrick also supports indexing using the Locality-Sensitive Hashing approach. This is an approximate nearest-neighbor search technique, and allows one to trade off similarity search time at the expense of accuracy. The index introduces two new tunable parameters:\n",
|
||||
"\n",
|
||||
"- The number of hyperplanes, which is provided as an argument to `create_lsh_index(num_hyperplanes)`. The more documents, the more hyperplanes are needed. LSH is a form of dimensionality reduction. The original embeddings are transformed into lower dimensional vectors where the number of components is the same as the number of hyperplanes.\n",
|
||||
"- The Hamming distance, an integer representing the breadth of the search. Smaller Hamming distances result in faster retreival but lower accuracy.\n",
|
||||
"- The Hamming distance, an integer representing the breadth of the search. Smaller Hamming distances result in faster retrieval but lower accuracy.\n",
|
||||
"\n",
|
||||
"Here's how you can create an index on the embeddings we loaded into Yellowbrick. We'll also re-run the previous chat session, but this time the retrieval will use the index. Note that for such a small number of documents, you won't see the benefit of indexing in terms of performance."
|
||||
]
|
||||
|
Loading…
Reference in New Issue
Block a user