{
  "fetched_at": "2026-10-19T07:26:39Z",
  "source_url": "https://www.kaggle.com/competitions/109-2-uta-dl-app-hw2/data",
  "pages": [
    {
      "name": "data-description",
      "content": "The files are briefly described below:\n\n<ol>\n<li> Code files\n<p> There are three iPython notebooks that show you how to train a model and create a prediction output in Kaggle format. Please run the code to understand the concepts, and modify them to build your own model. For functions of each file please read the \"README.txt\". The \"utility.py\" is used the demo code. </p>\n</li>\n\n<li> Data files\n<p>\n<ul>\n<li>\n\"harrison_features.npz\" - Pre-extracted features of images in harrison dataset. We extracted ImageNet and places365 CNN features.\n</li>\n<li>\n\"tag_list.txt\" - The list of hashtags for all image files.\n</li>\n</ul>\n</p>\n</li>\n</ol>",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Description",
      "content": "In Homework 2 we are going to train models that can recommend hashtags based on the content of an image. We will use the dataset \"HARRISON\". Please refer to the paper (https://arxiv.org/abs/1605.05054) for more details about the dataset.\n\nThe original data of HARRISON can be downloaded here: (https://drive.google.com/file/d/1B9NZf42J_GpslRNlTTAxvM5c9WRVl_1Z/view?usp=sharing). You can download the images and take a look, but this is not mandatory. We already extracted the image features for you and upload to Kaggle as \"harrison_features.npz\".\n\nThe demo code separates the training process into two stages: word2vec training and image-to-word2vec training. The notebook of last semester can be found here: https://www.kaggle.com/kuantinglai/train-devise\nYou can export and import the code to this competition.\n\nTo evaluate the model performance, the top 10 hashtags (KNN=10) recommended from your model will be used to calculate the hamming loss of ground truth.  Get your computer ready and have fun!",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Evaluation",
      "content": "We will use the Hamming loss function to evaluate the performance:\n<p> \\[ \\frac{1}{|Y|} \\sum_{i=1}^{|N|} \\sum_{j=1}^{|H_i|} XOR(y'_{ij}, y_{ij}) \\] </p>\n\n<p> where \\\\(  |N| \\\\) is the total number of images, \\\\( |Y| \\\\) is the total number of labels, \\\\(  |H_i| \\\\) is the number of labeled hashtags of image  \\\\( i \\\\). </p>\n<p> The \\\\(  y'_{ij} \\\\) is a binary output (True or False ) that predicts the existing of a hashtag  \\\\( h_{ij}  \\\\). The $y_{ij}$ is the ground truth. </p>",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "rules",
      "content": "This is a page where you can include rules that participants must accept before joining. You may wish to include rules like:\n\n - Don't cheat!\n - Apply yourself!\n - Have fun!",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    }
  ]
}