From 102b2fae4e5ce6397f6f754db71022bf8861e030 Mon Sep 17 00:00:00 2001
From: Nick Crews <nicholas.b.crews@gmail.com>
Date: Sun, 7 Jul 2024 17:25:29 -0800
Subject: [PATCH] docs: mention "precomputed" metric as a parameter

Inspired by https://github.com/lmcinnes/umap/issues/61
---
 doc/parameters.rst | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/doc/parameters.rst b/doc/parameters.rst
index 4fc60ca2..23d4c78a 100644
--- a/doc/parameters.rst
+++ b/doc/parameters.rst
@@ -341,6 +341,13 @@ of metrics, including:
 - sokalsneath
 - yule
 
+**Precomputed**
+
+All of the above metrics assume your input data is "raw" in some N-dimensional space.
+Sometimes, you have already calculated the pairwise distances between points,
+and your input data is a distance/similarity matrix.
+In this case, you can do something like ``UMAP(metric='precomputed').fit_transform(<distance matrix>)``.
+
 Any of which can be specified by setting ``metric='<metric name>'``; for
 example to use cosine distance as the metric you would use
 ``metric='cosine'``.