{
  "id": 547695,
  "title": "The problem of rounding 0.5",
  "url": "/competitions/child-mind-institute-problematic-internet-use/discussion/547695",
  "author_name": "",
  "post_date": "2024-11-23T04:47:27.535184800Z",
  "votes": 3,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Has anyone thought about whether 0.5 should be taken as an integer 1 or an integer 0 in weighted integration.</p>\n<p>I mean, when we use the weighted integration model, if the output number is not an integer but has a decimal like 0.5 0.6, how do we take an integer</p>",
  "messages": [
    {
      "id": "3053000",
      "postDate": "11/23/2024 04:47:27",
      "content": "<p>Has anyone thought about whether 0.5 should be taken as an integer 1 or an integer 0 in weighted integration.</p>\n<p>I mean, when we use the weighted integration model, if the output number is not an integer but has a decimal like 0.5 0.6, how do we take an integer</p>",
      "rawMarkdown": "Has anyone thought about whether 0.5 should be taken as an integer 1 or an integer 0 in weighted integration.\n\nI mean, when we use the weighted integration model, if the output number is not an integer but has a decimal like 0.5 0.6, how do we take an integer",
      "votes": null
    },
    {
      "id": "3053416",
      "postDate": "11/23/2024 13:05:42",
      "content": "<p>I'm not sure what you mean by 'weighted integration' but in my experience a model outputting exactly 0.5 is very rare if we're using 32 or 64-bit floats. So I'm not sure of the practical significance.</p>\n<p>You could add an if statement to your rounding function to always round 0.5 up or down, and see if changes the CV or LB score. I'd be surprised if it does, but please let us know if so - could be an interesting learning..</p>",
      "rawMarkdown": "I'm not sure what you mean by 'weighted integration' but in my experience a model outputting exactly 0.5 is very rare if we're using 32 or 64-bit floats. So I'm not sure of the practical significance.\n\nYou could add an if statement to your rounding function to always round 0.5 up or down, and see if changes the CV or LB score. I'd be surprised if it does, but please let us know if so - could be an interesting learning..",
      "votes": null
    },
    {
      "id": "3054263",
      "postDate": "11/24/2024 13:57:54",
      "content": "<p>Totally agree. Another interesting topic would be, what if sii lower than 0? I think this is not possible if you did a good job on data cleaning…However, I'm still curious about this www </p>",
      "rawMarkdown": "Totally agree. Another interesting topic would be, what if sii lower than 0? I think this is not possible if you did a good job on data cleaning...However, I'm still curious about this www",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3053416,
      "author_name": "josephmarturano",
      "author_url": "",
      "post_date": "11/23/2024 13:05:42",
      "content": "<p>I'm not sure what you mean by 'weighted integration' but in my experience a model outputting exactly 0.5 is very rare if we're using 32 or 64-bit floats. So I'm not sure of the practical significance.</p>\n<p>You could add an if statement to your rounding function to always round 0.5 up or down, and see if changes the CV or LB score. I'd be surprised if it does, but please let us know if so - could be an interesting learning..</p>",
      "votes": null,
      "replies": [
        {
          "id": 3054263,
          "author_name": "kaiyisung",
          "author_url": "",
          "post_date": "11/24/2024 13:57:54",
          "content": "<p>Totally agree. Another interesting topic would be, what if sii lower than 0? I think this is not possible if you did a good job on data cleaning…However, I'm still curious about this www </p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "3053000": "Has anyone thought about whether 0.5 should be taken as an integer 1 or an integer 0 in weighted integration.\n\nI mean, when we use the weighted integration model, if the output number is not an integer but has a decimal like 0.5 0.6, how do we take an integer",
    "3053416": "I'm not sure what you mean by 'weighted integration' but in my experience a model outputting exactly 0.5 is very rare if we're using 32 or 64-bit floats. So I'm not sure of the practical significance.\n\nYou could add an if statement to your rounding function to always round 0.5 up or down, and see if changes the CV or LB score. I'd be surprised if it does, but please let us know if so - could be an interesting learning..",
    "3054263": "Totally agree. Another interesting topic would be, what if sii lower than 0? I think this is not possible if you did a good job on data cleaning...However, I'm still curious about this www"
  },
  "source": "meta"
}