Fixed description text

Author: whitead

dl/xai.ipynb

@@ -755,7 +755,7 @@
     "from myst_nb import glue\n",
     "\n",
     "val = []\n",
-    "ms = np.linspace(2, 512, 25)\n",
+    "ms = np.linspace(2, 1024, 25)\n",
     "for m in ms.astype(np.int32):\n",
     "    sampled_x = train_x[:m]\n",
     "    val.append(\n",
@@ -856,7 +856,7 @@
    "source": [
     "As someone who works with peptides, I believe the Shapley values are the most accurate here. Shapley values tend to give a smoother, more balanced view of feature importance than raw gradients.\n",
     "\n",
-    "What can we conclude from this information? We could perhaps add an explanation like this: \"The sequence is predicted to be hemolytic primarily because of the tryptophan (W), tyrosine (Y), and lack of glycine (G).\"\n",
+    "What can we conclude from this information? We could perhaps add an explanation like this: \"The sequence is predicted to be hemolytic primarily because of the tryptophan (W) and tyrosine (Y) residues.\"\n",
     "\n",
     "## What is feature importance for?\n",
     "\n",