Update Mission244Solutions.ipynb

Mission244Solutions.ipynb

@@ -465,9 +465,9 @@
    "source": [
     "### Summary:\n",
     "\n",
-    "It looks like adding more neurons to the single hidden layer helped massively improved simple accuracy from approximately `86%` to approximately `94%`. Simple accuracy computes the number of correct classifications the model made, but doesn't tell us anything about false or true positives or false or true negatives.\n",
+    "It looks like adding more neurons to the single hidden layer improved simple accuracy to approximately `97%`. Simple accuracy computes the number of correct classifications the model made, but doesn't tell us anything about false or true positives or false or true negatives.\n",
     "\n",
-    "Given that k-nearest neighbors achieved approximately `96%` accuracy, there doesn't seem to be any advantages to using a single hidden layer neural network for this problem."
+    "Given that k-nearest neighbors achieved approximately `98%` accuracy, there doesn't seem to be any advantages to using a single hidden layer neural network for this problem."
    ]
   },
   {
@@ -549,7 +549,7 @@
    "source": [
     "### Summary\n",
     "\n",
-    "Using 2 hidden layers improved our simple accuracy to `95%`. While I'd traditionally be worried about overfitting, using 4-fold cross validation also gives me a bit more assurance that the model is generalizing to achieve the extra `1%` in simple accuracy over the single hidden layer networks we tried earlier."
+    "Using 2 hidden layers improved our simple accuracy to `98%`. While I'd traditionally be worried about overfitting, using 4-fold cross validation also gives me a bit more assurance that the model is generalizing to achieve the extra `1%` in simple accuracy over the single hidden layer networks we tried earlier."
    ]
   },
   {
@@ -698,7 +698,7 @@
    "source": [
     "### Summary\n",
     "\n",
-    "Using 3 hidden layers improved our simple accuracy to `96%`, even with 6-fold cross validation. This seems to be in line with the research literature out there about deep neural networks for computer vision. Having more layers and more neurons tends to improve the network's performance."
+    "Using 3 hidden layers returned a simple accuracy of nearly `98%`, even with 6-fold cross validation."
    ]
   }
  ],