Improved tutorial notebook

tutorials/pointcloud2hypergraph/pointnet_lifting.ipynb

@@ -42,7 +42,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "## Loading the Dataset - Wall Shear Stress on the Artery"
+    "## Loading the Dataset - Wall Shear Stress on the Artery Surface"
    ]
   },
   {
@@ -74,7 +74,9 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "Our dataset consists of 2000 triangular meshes which are concatenated and represented as simplicial complex. The `slices_pos` and `slices_face` can be used to extract individual meshes. The node attribute is the geodesic distance to the artery inlet (to give the mesh a direction) and the face attribute is the surface normal."
+    "Our dataset consists of 2000 triangular meshes which are concatenated and represented as simplicial complex. The `slices_pos` and `slices_face` can be used to extract individual meshes. The node attribute is the geodesic distance to the artery inlet (to give the mesh a direction) and the face attribute is the surface normal.\n",
+    "\n",
+    "Even though triangular meshes can be considered simplicial complices, they can be reduced to point clouds by dropping all `face` information. In order to demonstrate our lifting, we will do this here. We will also select a single mesh from the dataset to make things easier."
    ]
   },
   {
@@ -93,12 +95,15 @@
     "    x=dataset.x[slice_pos],\n",
     "    y=dataset.y[slice_pos],\n",
     "    pos=dataset.pos[slice_pos],\n",
-    "    face=dataset.face[:, slice_face],\n",
+    "    # face=dataset.face[:, slice_face],\n",
     "    # x_2=dataset.x_2[slice_face],\n",
     "    # incidence_2=dataset.incidence_2[slice_pos, slice_face]  # not supported by PyTorch,\n",
-    "    num_features=dataset.num_features,\n",
     ")\n",
     "\n",
+    "# Adapt dataset config\n",
+    "dataset_config[\"data_domain\"] = \"point_cloud\"\n",
+    "dataset_config[\"num_features\"] = 1\n",
+    "\n",
     "print(\"Data sample:\")\n",
     "data"
    ]