{
  "id": 209022,
  "title": "Why my test ensemble is too low!!!",
  "url": "/competitions/cassava-leaf-disease-classification/discussion/209022",
  "author_name": "SrihariMohan",
  "post_date": "2021-01-06T00:32:42.422000",
  "votes": 1,
  "comment_count": 4,
  "views": 0,
  "content": "<p>Not sure why this is happening. Running all the folds in training set is yielding me the validation accuracy of 0.86 approx. When I perform the same in test set, my accuracy is low to 0.139 :/ Is there anything I am missing to improve on this. Please suggest.</p>",
  "messages": [
    {
      "id": 1140362,
      "postDate": "2021-01-06T00:32:42.423Z",
      "content": "<p>Not sure why this is happening. Running all the folds in training set is yielding me the validation accuracy of 0.86 approx. When I perform the same in test set, my accuracy is low to 0.139 :/ Is there anything I am missing to improve on this. Please suggest.</p>",
      "rawMarkdown": "Not sure why this is happening. Running all the folds in training set is yielding me the validation accuracy of 0.86 approx. When I perform the same in test set, my accuracy is low to 0.139 :/ Is there anything I am missing to improve on this. Please suggest.\n\n",
      "votes": 1
    },
    {
      "id": 1145259,
      "postDate": "2021-01-09T02:30:20.230Z",
      "content": "<p>Thank you so much both. I think I made some mistake while performing ensemble. When I did very slowly, I got my expected result :-) </p>",
      "rawMarkdown": "Thank you so much both. I think I made some mistake while performing ensemble. When I did very slowly, I got my expected result :-) ",
      "replies": [
        {
          "id": 1153682,
          "postDate": "2021-01-15T04:57:42.637Z",
          "rawMarkdown": "",
          "isDeleted": true
        }
      ]
    },
    {
      "id": 1141129,
      "postDate": "2021-01-06T14:21:39.697Z",
      "content": "<p>Debug your inference method and check if you're averaging your predictions. I had the same problem, and it was due to a bad indentation in my inference method.</p>",
      "rawMarkdown": "Debug your inference method and check if you're averaging your predictions. I had the same problem, and it was due to a bad indentation in my inference method."
    },
    {
      "id": 1140412,
      "postDate": "2021-01-06T01:49:51.790Z",
      "content": "<p>Make sure you don't have a shuffle in the test set. You're basically getting random results. Run your inference against your train data to make sure you are not running out of memory or some other bug. </p>",
      "rawMarkdown": "Make sure you don't have a shuffle in the test set. You're basically getting random results. Run your inference against your train data to make sure you are not running out of memory or some other bug. "
    }
  ],
  "comments": [
    {
      "id": 1145259,
      "author_name": "SrihariMohan",
      "author_url": "",
      "post_date": "2021-01-09T02:30:20.230000",
      "content": "<p>Thank you so much both. I think I made some mistake while performing ensemble. When I did very slowly, I got my expected result :-) </p>",
      "votes": 0,
      "replies": [
        {
          "id": 1153682,
          "author_name": "",
          "author_url": "",
          "post_date": "2021-01-15T04:57:42.637000",
          "content": "",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1141129,
      "author_name": "Fares Lassoued",
      "author_url": "",
      "post_date": "2021-01-06T14:21:39.697000",
      "content": "<p>Debug your inference method and check if you're averaging your predictions. I had the same problem, and it was due to a bad indentation in my inference method.</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 1140412,
      "author_name": "quadcore/Richard Epstein",
      "author_url": "",
      "post_date": "2021-01-06T01:49:51.790000",
      "content": "<p>Make sure you don't have a shuffle in the test set. You're basically getting random results. Run your inference against your train data to make sure you are not running out of memory or some other bug. </p>",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1140362": "Not sure why this is happening. Running all the folds in training set is yielding me the validation accuracy of 0.86 approx. When I perform the same in test set, my accuracy is low to 0.139 :/ Is there anything I am missing to improve on this. Please suggest.\n\n",
    "1145259": "Thank you so much both. I think I made some mistake while performing ensemble. When I did very slowly, I got my expected result :-) ",
    "1141129": "Debug your inference method and check if you're averaging your predictions. I had the same problem, and it was due to a bad indentation in my inference method.",
    "1140412": "Make sure you don't have a shuffle in the test set. You're basically getting random results. Run your inference against your train data to make sure you are not running out of memory or some other bug. "
  }
}