Add net diagrams

FFNN_RNN.ipynb

@@ -1277,6 +1277,13 @@
     "model.summary()"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img src='./images/dnnTutorial.png'>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -3775,7 +3782,16 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "Let's look at a fancier way of solving the same classification problem. In this case we will use Recurrent Neural Netwroks. These allow you to process variable length sequences of data. For example, we can use them to describe an event in terms of the properties of its jets, whose number varies event by event. We could also describe the same event using the properties of its muons, or any other particle that appears in it. Because the number of particles of each type changes in each event, we need the flexibility of RNNs to process this type of data."
+    "Let's look at a fancier way of solving the same classification problem. In this case we will use Recurrent Neural Netwroks. These allow you to process variable length sequences of data. For example, we can use them to describe an event in terms of the properties of its jets, whose number varies event by event. We could also describe the same event using the properties of its muons, or any other particle that appears in it. Because the number of particles of each type changes in each event, we need the flexibility of RNNs to process this type of data.\n",
+    "<img src='./images/RNNdetail.png'>"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "We will build an event-level classifier with 4 recurrent branches for the 4 different types of particles in the event:\n",
+    "<img src='./images/rnnTutorial.png'>"
    ]
   },
   {