{
  "id": 109726,
  "title": "Sugar baseline",
  "url": "/competitions/understanding_cloud_organization/discussion/109726",
  "author_name": "",
  "post_date": "2019-09-21T18:15:27.316587400Z",
  "votes": 3,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Predicting all pixels as \"sugar\" yields a score of 0.393, see notebook below. Has anybody tried the other classes? If not it could be done using a variation of my code. Might be interesting to know about distribution of classes between training and test set.</p>\n\n<p><a href=\"https://www.kaggle.com/janlauge/cloud-eda-and-single-class-prediction-baseline\">https://www.kaggle.com/janlauge/cloud-eda-and-single-class-prediction-baseline</a></p>",
  "messages": [
    {
      "id": "631267",
      "postDate": "09/21/2019 18:15:27",
      "content": "<p>Predicting all pixels as \"sugar\" yields a score of 0.393, see notebook below. Has anybody tried the other classes? If not it could be done using a variation of my code. Might be interesting to know about distribution of classes between training and test set.</p>\n\n<p><a href=\"https://www.kaggle.com/janlauge/cloud-eda-and-single-class-prediction-baseline\">https://www.kaggle.com/janlauge/cloud-eda-and-single-class-prediction-baseline</a></p>",
      "rawMarkdown": "Predicting all pixels as \"sugar\" yields a score of 0.393, see notebook below. Has anybody tried the other classes? If not it could be done using a variation of my code. Might be interesting to know about distribution of classes between training and test set.\n\nhttps://www.kaggle.com/janlauge/cloud-eda-and-single-class-prediction-baseline",
      "votes": null
    },
    {
      "id": "641769",
      "postDate": "10/05/2019 04:21:35",
      "content": "<p>is it good to have four binary segmentation problems from original dataset and then just do predictons on the four of them denoting  all 4 classes</p>",
      "rawMarkdown": "is it good to have four binary segmentation problems from original dataset and then just do predictons on the four of them denoting  all 4 classes",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 641769,
      "author_name": "pranshu29",
      "author_url": "",
      "post_date": "10/05/2019 04:21:35",
      "content": "<p>is it good to have four binary segmentation problems from original dataset and then just do predictons on the four of them denoting  all 4 classes</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "631267": "Predicting all pixels as \"sugar\" yields a score of 0.393, see notebook below. Has anybody tried the other classes? If not it could be done using a variation of my code. Might be interesting to know about distribution of classes between training and test set.\n\nhttps://www.kaggle.com/janlauge/cloud-eda-and-single-class-prediction-baseline",
    "641769": "is it good to have four binary segmentation problems from original dataset and then just do predictons on the four of them denoting  all 4 classes"
  },
  "source": "meta"
}