{
  "fetched_at": "2026-10-19T18:56:49Z",
  "source_url": "https://www.kaggle.com/competitions/ell409-assignment-1/data",
  "pages": [
    {
      "name": "rules",
      "content": " - Don't use standard libraries like keras, pytorch, tensorflow etc. for implementing the neural net.\n - Do it on individual basis.\n - Don't cheat!\n - Apply yourself!\n - Have fun!",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "data-description",
      "content": "<p>This page shows up underneath the data files. Here you describe what files you have provided and the format of each. There is no single format for this page that is appropriate for all competitons, but you should strive to describe as much as you can here. A little time spent describing the data here can save a lot of time answering questions later.</p>\n<p>Participants should be able to answer questions like these after reading the data description: What files do I need? What should I expect the data format to be? What am I predicting? What acroynms will I encounter?</p>\n<h2>File descriptions</h2>\n<ul>\n<li><strong>mnist_train.csv</strong> - the training set</li>\n<li><strong>mnist_test.csv</strong> - the test set</li>\n<li><strong>sampleSubmission.csv</strong> - a sample submission file in the correct format</li>\n</ul>\n<h2>Train File</h2>\n<p> The train file consist of first column as the Labels of the digit. The 28x28 image of digits has been flattened into an array of 1x784 and appended next to the label, making the first entry of 1x785. \nThe test only consists of flattened pixels and no labels. So the dimension of row will be 1x784 corresponding to 28x28 pixels of some digit. \n</ul>",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Description",
      "content": "The MNIST database is a large database of handwritten digits that is commonly used for training various image processing systems. The database is also widely used for training and testing in the field of machine learning.\n\nYou have to make a multiclass neural network classifier that classifies each image as corresponding digit. Your code should be flexible enough so that it can model any specification that will be given to you during the demo of assignment. E.g. You could be asked to make a neural network of 2 layers(1 input + 1 hidden + 1 output, input layer is not counted as a separate layer and specified number of nodes(neurons) in the hidden layers say 50) with certain initial weights and then you could be asked to give the values of weight/output after ,say, 3 iteration of some given input. This will ensure correctness of your code even if your accuracy is low.\n\nThe weightage of the assignment to the course will be decided later. You will be evaluated on accuracy(40%), correctness(60%).\n\nPlagarism will be heavily penalized. You are not allowed to use nerual network libraries like keras. You can also use matlab but again don't use any built in functions related to neural network.\n\nThis assignment is to be done on individual basis.",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Evaluation",
      "content": "<p>The evaluation pages describes how submissions will be scored and how students should format their submissions. You don't need a sub-title at the top (the page title appears above). For an example of a high quality evaluation page, see the 2013 KDD Cup page, which we have copied here:</p>\r\n<hr />\r\n<p>The evaluation metric for this competition is <a href=\"https://www.kaggle.com/wiki/MeanFScore\"> Mean F1-Score</a>. The F1 score, commonly used in information retrieval, measures accuracy using the statistics precision p and recall r. Precision is the ratio of true positives (tp) to all predicted positives (tp + fp). Recall is the ratio of true positives to all actual positives (tp + fn). The F1 score is given by:</p>\r\n<p>\\[ F1 = 2\\frac{p \\cdot r}{p+r}\\ \\ \\mathrm{where}\\ \\ p = \\frac{tp}{tp+fp},\\ \\ r = \\frac{tp}{tp+fn} \\]</p>\r\n<p>The F1 metric weights recall and precision equally, and a good retrieval algorithm will maximize both precision and recall simultaneously. Thus, moderately good performance on both will be favored over extremely good performance on one and poor performance on the other.</p>\r\n<h2>Submission Format</h2>\r\n<p><strong>For every author in the dataset</strong>, submission files should contain two columns: AuthorId and DuplicateAuthorIds. DuplicateAuthorIds should be a space-delimited list. Every AuthorId counts as his/her own duplicate, and every duplicate should be listed under each of its respective ids. For example, if you suspect author A, B, and C are the same, you should list (A,A B C), (B,B A C), (C,C A B).</p>\r\n<p>The file should contain a header and have the following format:</p>\r\n<pre>AuthorId,DuplicateAuthorIds<br />1,1<br /><span style=\"font-size: 14px; line-height: 1.4em;\">8,8<br />9,9 10<br /></span><span style=\"font-size: 14px; line-height: 1.4em;\">10,10 9<br /></span><span style=\"font-size: 14px; line-height: 1.4em;\">etc.</span></pre>",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    }
  ]
}