{
  "id": 321824,
  "title": "What's a good value of RMSE for this problem?",
  "url": "/competitions/denoising-shabby-pages/discussion/321824",
  "author_name": "",
  "post_date": "2022-04-28T21:05:06.240852300Z",
  "votes": 1,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hi! Thanks for this competition. This is a very cool project.</p>\n<p>I'm curious about of what would be a good value for RMSE in practice.<br>\nWould be useful in practice a model that could achieve a RMSE &lt;= 0.2 in test set?</p>\n<p>Do you have a specific value to serve as a baseline?<br>\nThanks!</p>",
  "messages": [
    {
      "id": "1771034",
      "postDate": "04/28/2022 21:05:06",
      "content": "<p>Hi! Thanks for this competition. This is a very cool project.</p>\n<p>I'm curious about of what would be a good value for RMSE in practice.<br>\nWould be useful in practice a model that could achieve a RMSE &lt;= 0.2 in test set?</p>\n<p>Do you have a specific value to serve as a baseline?<br>\nThanks!</p>",
      "rawMarkdown": "Hi! Thanks for this competition. This is a very cool project.\n\nI'm curious about of what would be a good value for RMSE in practice.\nWould be useful in practice a model that could achieve a RMSE <= 0.2 in test set?\n\nDo you have a specific value to serve as a baseline?\nThanks!",
      "votes": null
    },
    {
      "id": "1776247",
      "postDate": "05/03/2022 20:35:37",
      "content": "<p>The <a href=\"https://www.kaggle.com/competitions/denoising-dirty-documents\" target=\"_blank\">Denoising Dirty Documents</a> leaderboard (now 7 years old) serves as a reference point for what one might be able to achieve.  That data set is less random, so it is going to be easier to reach smaller RMSE values. That said, many techniques available today were not available then. So while it's hard to make a guess just yet, we can assume that over time the error levels will get to be less than 5%. However, we are not at that stage of the competition yet. Therefore, for now, an error level of less than 20% would be ideal. Hope this helps!</p>",
      "rawMarkdown": "The [Denoising Dirty Documents](https://www.kaggle.com/competitions/denoising-dirty-documents) leaderboard (now 7 years old) serves as a reference point for what one might be able to achieve.  That data set is less random, so it is going to be easier to reach smaller RMSE values. That said, many techniques available today were not available then. So while it's hard to make a guess just yet, we can assume that over time the error levels will get to be less than 5%. However, we are not at that stage of the competition yet. Therefore, for now, an error level of less than 20% would be ideal. Hope this helps!",
      "votes": null
    },
    {
      "id": "1799057",
      "postDate": "05/23/2022 15:54:18",
      "content": "<p>What approach are you using?</p>",
      "rawMarkdown": "What approach are you using?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1776247,
      "author_name": "samaymaini",
      "author_url": "",
      "post_date": "05/03/2022 20:35:37",
      "content": "<p>The <a href=\"https://www.kaggle.com/competitions/denoising-dirty-documents\" target=\"_blank\">Denoising Dirty Documents</a> leaderboard (now 7 years old) serves as a reference point for what one might be able to achieve.  That data set is less random, so it is going to be easier to reach smaller RMSE values. That said, many techniques available today were not available then. So while it's hard to make a guess just yet, we can assume that over time the error levels will get to be less than 5%. However, we are not at that stage of the competition yet. Therefore, for now, an error level of less than 20% would be ideal. Hope this helps!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1799057,
      "author_name": "deepdreamx",
      "author_url": "",
      "post_date": "05/23/2022 15:54:18",
      "content": "<p>What approach are you using?</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1771034": "Hi! Thanks for this competition. This is a very cool project.\n\nI'm curious about of what would be a good value for RMSE in practice.\nWould be useful in practice a model that could achieve a RMSE <= 0.2 in test set?\n\nDo you have a specific value to serve as a baseline?\nThanks!",
    "1776247": "The [Denoising Dirty Documents](https://www.kaggle.com/competitions/denoising-dirty-documents) leaderboard (now 7 years old) serves as a reference point for what one might be able to achieve.  That data set is less random, so it is going to be easier to reach smaller RMSE values. That said, many techniques available today were not available then. So while it's hard to make a guess just yet, we can assume that over time the error levels will get to be less than 5%. However, we are not at that stage of the competition yet. Therefore, for now, an error level of less than 20% would be ideal. Hope this helps!",
    "1799057": "What approach are you using?"
  },
  "source": "meta"
}