diff --git a/examples/heteroskedastic.ipynb b/examples/heteroskedastic.ipynb
index 6b1e806..94eb281 100644
--- a/examples/heteroskedastic.ipynb
+++ b/examples/heteroskedastic.ipynb
@@ -1,5 +1,28 @@
 {
  "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# Heteroskedastic fully and partially Bayesian neural nets\n",
+    "\n",
+    "*Prepared by Maxim Ziatdinov (October 2024)*"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "We usually assume by default a constant observation noise across all inputs. However, this assumption doesn't always hold in real-world datasets which may exhibit input-dependent levels of noise. NeuroBayes offers heteroskedastic BNNs that can capture varying levels of noise in different regions of the data, allowing for more accurate uncertainty quantification."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Imports:"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": 1,