diff --git a/docs/notebooks/multivariate_gaussian_distribution.ipynb b/docs/notebooks/multivariate_gaussian_distribution.ipynb
index 07eb009..bb77e05 100644
--- a/docs/notebooks/multivariate_gaussian_distribution.ipynb
+++ b/docs/notebooks/multivariate_gaussian_distribution.ipynb
@@ -3,7 +3,7 @@
   {
    "cell_type": "code",
    "execution_count": 1,
-   "id": "bb9ce462",
+   "id": "cf172a64",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -18,7 +18,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "a2f0fbbf",
+   "id": "72cca938",
    "metadata": {},
    "source": [
     "In this notebook, we sample a 2d gaussian posterior using LINNA. LINNA isn't designed for low dimension posteiors, so the performance will not be great. However, this notebook illustrates how one can use LINNA to sample posteriors. "
@@ -26,7 +26,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "10743380",
+   "id": "ae184c0a",
    "metadata": {},
    "source": [
     "# Create a multivariate gaussian distribution "
@@ -35,7 +35,7 @@
   {
    "cell_type": "code",
    "execution_count": 2,
-   "id": "a4a8c4fa",
+   "id": "c0893db6",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -65,7 +65,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "e80d182a",
+   "id": "4e936434",
    "metadata": {},
    "source": [
     "# Perform MCMC sampling using Linna"
@@ -74,7 +74,7 @@
   {
    "cell_type": "code",
    "execution_count": 3,
-   "id": "86cf1a71",
+   "id": "19103732",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -85,7 +85,7 @@
   {
    "cell_type": "code",
    "execution_count": 4,
-   "id": "a688cd37",
+   "id": "27d1fd82",
    "metadata": {},
    "outputs": [
     {
@@ -204,7 +204,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "373abca6",
+   "id": "f004d94b",
    "metadata": {},
    "source": [
     "# Check the result"
@@ -213,7 +213,7 @@
   {
    "cell_type": "code",
    "execution_count": 5,
-   "id": "0c50186c",
+   "id": "cee5428f",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -223,7 +223,7 @@
   {
    "cell_type": "code",
    "execution_count": 6,
-   "id": "b63b40bd",
+   "id": "1c673921",
    "metadata": {},
    "outputs": [
     {
@@ -259,7 +259,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "4ea71bed",
+   "id": "831779a5",
    "metadata": {},
    "source": [
     "# Useful performance check tools"
@@ -267,7 +267,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "8572b276",
+   "id": "d4b3bec9",
    "metadata": {},
    "source": [
     "The output of linna has the following struture."
@@ -276,7 +276,7 @@
   {
    "cell_type": "code",
    "execution_count": 2,
-   "id": "940aab44",
+   "id": "4ee3408e",
    "metadata": {},
    "outputs": [
     {
@@ -384,7 +384,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "ca9a4769",
+   "id": "2a0fc5d9",
    "metadata": {},
    "source": [
     "In each iteration, \n",
@@ -402,16 +402,25 @@
   },
   {
    "cell_type": "markdown",
-   "id": "2545f502",
+   "id": "ef4a2850",
    "metadata": {},
    "source": [
     "Note that if your job crashs at an iteration, LINNA can be restarted from the previous iteration by cleanining the directories corresponding to the crashed iteration and rerunning the code. "
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "aa39713c",
+   "metadata": {},
+   "source": [
+    "#### To retrieve the model \n",
+    "One might wish to use the learned model to perform fast model evaluation. This can be done with the following functions. "
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": 3,
-   "id": "05b5ecc3",
+   "id": "df72d84a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -421,7 +430,7 @@
   {
    "cell_type": "code",
    "execution_count": 4,
-   "id": "2c1d75d8",
+   "id": "a32c8eb4",
    "metadata": {
     "scrolled": true
    },
@@ -442,7 +451,7 @@
   {
    "cell_type": "code",
    "execution_count": 7,
-   "id": "0ef1c004",
+   "id": "64cd2284",
    "metadata": {},
    "outputs": [
     {