From 1d3eac71d84f88dbc169c143d50f659983171636 Mon Sep 17 00:00:00 2001 From: Alex Klibisz Date: Wed, 18 Aug 2021 23:00:36 -0400 Subject: [PATCH] Docs: fixed error in Tour de Elastiknn post --- docs/_posts/2021-07-30-how-does-elastiknn-work.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/docs/_posts/2021-07-30-how-does-elastiknn-work.md b/docs/_posts/2021-07-30-how-does-elastiknn-work.md index 65b6e8cae..6d7202dd6 100644 --- a/docs/_posts/2021-07-30-how-does-elastiknn-work.md +++ b/docs/_posts/2021-07-30-how-does-elastiknn-work.md @@ -228,7 +228,7 @@ This hash function is defined more precisely as: $$h_j(v) = \lfloor \frac{A_j \cdot v + B_j}{w} \rfloor$$ -Each vector $$A_j$$ is sampled from a standard Normal distribution and each scalar $$B_j$$ is sampled from a uniform distribution in $$[0, 1]$$. +Each vector $$A_j$$ is sampled from a standard Normal distribution and each scalar $$B_j$$ is sampled from a uniform distribution in $$[0, w]$$. The width $$w$$ is provided as a hyper-parameter and depends on the magnitude of the vectors. As vector values increase, the scalar projections will increase, so $$w$$ should also increase.