From 57eeed8687118cf39b0985f1b9fb52578292156d Mon Sep 17 00:00:00 2001
From: Emile Mathieu <emile.mathieu@stats.ox.ac.uk>
Date: Fri, 5 Apr 2024 12:08:37 +0100
Subject: [PATCH] add sentences on diffusion at the end

---
 _posts/2024-01-20-flow-matching.md | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/_posts/2024-01-20-flow-matching.md b/_posts/2024-01-20-flow-matching.md
index 7c55541..91d1a50 100644
--- a/_posts/2024-01-20-flow-matching.md
+++ b/_posts/2024-01-20-flow-matching.md
@@ -1720,6 +1720,9 @@ Similarly to CNFs, sampled can be obtained at inference time by solving the ODE
 
 <!-- which bridge between a noise refence distribution and a target data distribution, and as such constructs a generative model. -->
 
+In this post we have not talked about diffusion (i.e. score based) models on purpose as they are not necessary for understanding flow matching.
+Yet these are deeply related and even exactly the same in some setting.
+We are planning to explore these connections, along with generalisations in a follow-up post!
 
 # Citation