Update mnist.html

mnist.html

@@ -40,7 +40,7 @@
 						<!-- Post -->
 							<section class="post">
 								<header class="major">
-									<h1>COMPARING KNN, NN AND CNN USING MNIST AND FASHION_MNIST DATASETS</h1>
+									<h1>COMPARING KNN, NN AND CNN USING MNIST AND Fashion-MNIST DATASETS</h1>
 									<p><br /></p>
 									<a href="https://github.com/theo-xiao-sg/mnist_recognition" target="_blank" class="button">Source code</a>
 									<p><br /></p>
@@ -51,7 +51,7 @@ <h2>Data</h2>
 									<img src="images/mnist_illustration.jpg" style="width: 70%; height: auto;">
 									<p>The <b>Fashion-MNIST dataset</b> is a collection of images of fashion items, like T-shirts, dresses, shoes etc (as shown below)(<a href="https://en.wikipedia.org/wiki/Fashion_MNIST" style="color: blue;">link</a>). Since it represents much more complex computer vision task, many researchers use it as a replacement for the original MNIST digits dataset. It shares the same image size (28x28 pixels and grayscale) and has 60,000 training and 10,000 testing images. </p>
 									<img src="images/fashion_mnist_illustration.jpg" style="width: 70%; height: auto;">
-									<p>Fashion_MNIST dataset has 10 categories of output labels: 0: T-shirt/top, 1: Trouser, 2: Pullover, 3: Dress, 4: Coat, 5: Sandal, 6: Shirt, 7: Sneaker, 8: Bag, 9: Ankle Boot.</p>
+									<p>Fashion-MNIST dataset has 10 categories of output labels: 0: T-shirt/top, 1: Trouser, 2: Pullover, 3: Dress, 4: Coat, 5: Sandal, 6: Shirt, 7: Sneaker, 8: Bag, 9: Ankle Boot.</p>
 									<h2>Results</h2>
 									<h3>MNIST dataset Results</h3>
 									<p>I compared three models: KNN, Neural Network (NN), and Convolutional Neural Network (CNN), by classifying the MNIST dataset. I stated the accuracies on the testing dataset for all the three models as follows:</p>
@@ -93,7 +93,7 @@ <h3>c. CNN model:</h3>
 </code></pre>
 								<p>For MNIST dataset, KNN had accuracy of 96.59% -97.05%, and Neural network model achieved 96.85% - 97.85%.  As expected, CNN model achieved the highest accuracy of 99.06% - 99.28% among all three models. </p>
 							<h3>Fashion-MNIST dataset Results</h3>
-							<p>I compared three models: KNN, Neural Network (NN), and Convolutional Neural Network (CNN), by classifying the Fashion_MNIST dataset. I stated the accuracy on the testing dataset for all three models as below:</p>
+							<p>I compared three models: KNN, Neural Network (NN), and Convolutional Neural Network (CNN), by classifying the Fashion-MNIST dataset. I stated the accuracy on the testing dataset for all three models as below:</p>
 							<h3>a. KKN model:</h3>
 								<!-- Preformatted Code -->
 								<style>
@@ -132,7 +132,7 @@ <h3>c. CNN model:</h3>
 CNN with 4 hidden convolutional layers;	accuracy: 91.40%
 CNN with 4 hidden convolutional layers and dropout layers;	accuracy: 92.14%
 </code></pre>
-								<p>After comparing these results of Fashion_MNIST dataset with the results of MNIST dataset, the Fashion_MNIST dataset was indeed a lot more challenging to all the three models tested here. The accuracies on the testing data dropped for all three models. KNN had accuracy of 84.97% -85.54%, and Neural network model achieved 88.68% - 89.83%.  Still, CNN model achieved the highest accuracy of 91.18% - 92.14% among all three models.</p>
+								<p>After comparing these results of Fashion-MNIST dataset with the results of MNIST dataset, the Fashion-MNIST dataset was indeed a lot more challenging to all the three models tested here. The accuracies on the testing data dropped for all three models. KNN had accuracy of 84.97% -85.54%, and Neural network model achieved 88.68% - 89.83%.  Still, CNN model achieved the highest accuracy of 91.18% - 92.14% among all three models.</p>
 							</section>
 					</div>
 				<!-- Footer -->