{
  "id": 534775,
  "title": "Comparison of 3 CNNs | 99.310% on Test dataset",
  "url": "/competitions/digit-recognizer/discussion/534775",
  "author_name": "Angelo Monteiro",
  "post_date": "2024-09-18T11:11:46.654000",
  "votes": 3,
  "comment_count": 1,
  "views": null,
  "content": "<p>Hi All,<br>\nI trained 3 CNNs because I wanted to check if running a pre-trained CNN on MNIST would make a big difference in terms of accuracy on the test dataset. </p>\n<p>It turns out that somethings bigger is not better. Check it out and do provide feedback if you want.</p>\n<p><a href=\"https://www.kaggle.com/code/angelogg/mnist-99-310-3-cnns-comparison\" target=\"_blank\">https://www.kaggle.com/code/angelogg/mnist-99-310-3-cnns-comparison</a></p>",
  "messages": [
    {
      "id": 2992231,
      "postDate": "2024-09-18T11:11:46.653Z",
      "content": "<p>Hi All,<br>\nI trained 3 CNNs because I wanted to check if running a pre-trained CNN on MNIST would make a big difference in terms of accuracy on the test dataset. </p>\n<p>It turns out that somethings bigger is not better. Check it out and do provide feedback if you want.</p>\n<p><a href=\"https://www.kaggle.com/code/angelogg/mnist-99-310-3-cnns-comparison\" target=\"_blank\">https://www.kaggle.com/code/angelogg/mnist-99-310-3-cnns-comparison</a></p>",
      "rawMarkdown": "Hi All,\nI trained 3 CNNs because I wanted to check if running a pre-trained CNN on MNIST would make a big difference in terms of accuracy on the test dataset. \n\nIt turns out that somethings bigger is not better. Check it out and do provide feedback if you want.\n\nhttps://www.kaggle.com/code/angelogg/mnist-99-310-3-cnns-comparison",
      "votes": 3
    },
    {
      "id": 2993593,
      "postDate": "2024-09-20T00:51:22.470Z",
      "content": "<p>You are right, bigger isn't always better for this case! <br>\nAn Architecture that allows building deep models is a Resnet by skipping some layers allowing the gradient to propagate more efficiently.</p>",
      "rawMarkdown": "You are right, bigger isn't always better for this case! \nAn Architecture that allows building deep models is a Resnet by skipping some layers allowing the gradient to propagate more efficiently."
    }
  ],
  "comments": [
    {
      "id": 2993593,
      "author_name": "Anas Hassar",
      "author_url": "",
      "post_date": "2024-09-20T00:51:22.470000",
      "content": "<p>You are right, bigger isn't always better for this case! <br>\nAn Architecture that allows building deep models is a Resnet by skipping some layers allowing the gradient to propagate more efficiently.</p>",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2992231": "Hi All,\nI trained 3 CNNs because I wanted to check if running a pre-trained CNN on MNIST would make a big difference in terms of accuracy on the test dataset. \n\nIt turns out that somethings bigger is not better. Check it out and do provide feedback if you want.\n\nhttps://www.kaggle.com/code/angelogg/mnist-99-310-3-cnns-comparison",
    "2993593": "You are right, bigger isn't always better for this case! \nAn Architecture that allows building deep models is a Resnet by skipping some layers allowing the gradient to propagate more efficiently."
  }
}