{
  "id": 551711,
  "title": "I wonder...",
  "url": "/competitions/child-mind-institute-problematic-internet-use/discussion/551711",
  "author_name": "",
  "post_date": "2024-12-14T22:55:28.716621600Z",
  "votes": 1,
  "comment_count": 5,
  "views": 0,
  "content": "<p>I tried changing every parameter and slighlty different architectures . But can't beat LB - little change and I drop down hard. Is it really that perfect or am I missing something?<br>\nBecause everything has already been tried here?<br>\nSo what else can help?<br>\nSorry for that question</p>",
  "messages": [
    {
      "id": "3072243",
      "postDate": "12/14/2024 22:55:28",
      "content": "<p>I tried changing every parameter and slighlty different architectures . But can't beat LB - little change and I drop down hard. Is it really that perfect or am I missing something?<br>\nBecause everything has already been tried here?<br>\nSo what else can help?<br>\nSorry for that question</p>",
      "rawMarkdown": "I tried changing every parameter and slighlty different architectures . But can't beat LB - little change and I drop down hard. Is it really that perfect or am I missing something?\nBecause everything has already been tried here?\nSo what else can help?\nSorry for that question",
      "votes": null
    },
    {
      "id": "3072258",
      "postDate": "12/15/2024 00:24:07",
      "content": "<p>This competition feels like a complete game of luck. I’ve done many experiments in the early stages, but they didn’t provide any real advantage. Models with good CV performance are behaving poorly on LB, and even a slight change in the random seed worse alters the results. Everything just feels chaotic and unpredictable. </p>\n<p>So, just wait for the competition to end and see what will Happen</p>",
      "rawMarkdown": "This competition feels like a complete game of luck. I’ve done many experiments in the early stages, but they didn’t provide any real advantage. Models with good CV performance are behaving poorly on LB, and even a slight change in the random seed worse alters the results. Everything just feels chaotic and unpredictable. \n\nSo, just wait for the competition to end and see what will Happen",
      "votes": null
    },
    {
      "id": "3072469",
      "postDate": "12/15/2024 08:51:22",
      "content": "<blockquote>\n  <p>This competition feels like a complete game of luck</p>\n</blockquote>\n<p>This is indeed a test of luck and lottery <a href=\"https://www.kaggle.com/abdmental01\" target=\"_blank\">@abdmental01</a> </p>",
      "rawMarkdown": "> This competition feels like a complete game of luck\n\nThis is indeed a test of luck and lottery @abdmental01",
      "votes": null
    },
    {
      "id": "3073760",
      "postDate": "12/16/2024 21:31:14",
      "content": "<blockquote>\n  <p>This competition feels like a complete game of luck</p>\n  <p>This is indeed a test of luck and lottery</p>\n</blockquote>\n<p>I quit this chaos 2 months ago haha!!<br>\nedit: 3 months ago*</p>",
      "rawMarkdown": "> This competition feels like a complete game of luck\n\n> This is indeed a test of luck and lottery\n\nI quit this chaos 2 months ago haha!!\nedit: 3 months ago*",
      "votes": null
    },
    {
      "id": "3073916",
      "postDate": "12/17/2024 05:16:43",
      "content": "<p>I think this is a good decision <a href=\"https://www.kaggle.com/eraakash\" target=\"_blank\">@eraakash</a> <br>\nThe same thing happened with ICR also, where people with 2-3 submissions performed better than ones with so-called optimization endeavors</p>",
      "rawMarkdown": "I think this is a good decision @eraakash \nThe same thing happened with ICR also, where people with 2-3 submissions performed better than ones with so-called optimization endeavors",
      "votes": null
    },
    {
      "id": "3073995",
      "postDate": "12/17/2024 07:38:17",
      "content": "<p>I can agree that it's a good decision. But I did the stupidest thing I could imagine. When I looked back at my 3mo old code, … </p>\n<ul>\n<li><p>I dropped all those columns in train that were not present in test (except target sii) </p></li>\n<li><p>I treated that as a regression problem and rounded the predictions to obtain integers</p></li>\n<li><p>I was optimizing RMSE.. But there's more - </p></li>\n</ul>\n<p>I tried to \"revamp\" that old approach by all the things I learnt in 3 months but woah, later when I checked the competition code,… The evaluation metric (whose name i never heard before) and basically the objective was entirely different! </p>\n<p>The 'comeback' just backfired so bad for trusting my old self haha!! Maybe aligning my approach with the correct objective can improve my score in the remaining time..? Too late I guess!</p>\n<p>But, as an active competitor, what are your thoughts on the final results of this competition? Will there be a massive shakeup? </p>",
      "rawMarkdown": "I can agree that it's a good decision. But I did the stupidest thing I could imagine. When I looked back at my 3mo old code, ... \n\n- I dropped all those columns in train that were not present in test (except target sii) \n\n- I treated that as a regression problem and rounded the predictions to obtain integers\n\n- I was optimizing RMSE.. But there's more - \n\nI tried to \"revamp\" that old approach by all the things I learnt in 3 months but woah, later when I checked the competition code,... The evaluation metric (whose name i never heard before) and basically the objective was entirely different! \n\nThe 'comeback' just backfired so bad for trusting my old self haha!! Maybe aligning my approach with the correct objective can improve my score in the remaining time..? Too late I guess!\n\nBut, as an active competitor, what are your thoughts on the final results of this competition? Will there be a massive shakeup?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3072258,
      "author_name": "abdmental01",
      "author_url": "",
      "post_date": "12/15/2024 00:24:07",
      "content": "<p>This competition feels like a complete game of luck. I’ve done many experiments in the early stages, but they didn’t provide any real advantage. Models with good CV performance are behaving poorly on LB, and even a slight change in the random seed worse alters the results. Everything just feels chaotic and unpredictable. </p>\n<p>So, just wait for the competition to end and see what will Happen</p>",
      "votes": null,
      "replies": [
        {
          "id": 3072469,
          "author_name": "ravi20076",
          "author_url": "",
          "post_date": "12/15/2024 08:51:22",
          "content": "<blockquote>\n  <p>This competition feels like a complete game of luck</p>\n</blockquote>\n<p>This is indeed a test of luck and lottery <a href=\"https://www.kaggle.com/abdmental01\" target=\"_blank\">@abdmental01</a> </p>",
          "votes": null,
          "replies": [
            {
              "id": 3073760,
              "author_name": "eraakash",
              "author_url": "",
              "post_date": "12/16/2024 21:31:14",
              "content": "<blockquote>\n  <p>This competition feels like a complete game of luck</p>\n  <p>This is indeed a test of luck and lottery</p>\n</blockquote>\n<p>I quit this chaos 2 months ago haha!!<br>\nedit: 3 months ago*</p>",
              "votes": null,
              "replies": [
                {
                  "id": 3073916,
                  "author_name": "ravi20076",
                  "author_url": "",
                  "post_date": "12/17/2024 05:16:43",
                  "content": "<p>I think this is a good decision <a href=\"https://www.kaggle.com/eraakash\" target=\"_blank\">@eraakash</a> <br>\nThe same thing happened with ICR also, where people with 2-3 submissions performed better than ones with so-called optimization endeavors</p>",
                  "votes": null,
                  "replies": [
                    {
                      "id": 3073995,
                      "author_name": "eraakash",
                      "author_url": "",
                      "post_date": "12/17/2024 07:38:17",
                      "content": "<p>I can agree that it's a good decision. But I did the stupidest thing I could imagine. When I looked back at my 3mo old code, … </p>\n<ul>\n<li><p>I dropped all those columns in train that were not present in test (except target sii) </p></li>\n<li><p>I treated that as a regression problem and rounded the predictions to obtain integers</p></li>\n<li><p>I was optimizing RMSE.. But there's more - </p></li>\n</ul>\n<p>I tried to \"revamp\" that old approach by all the things I learnt in 3 months but woah, later when I checked the competition code,… The evaluation metric (whose name i never heard before) and basically the objective was entirely different! </p>\n<p>The 'comeback' just backfired so bad for trusting my old self haha!! Maybe aligning my approach with the correct objective can improve my score in the remaining time..? Too late I guess!</p>\n<p>But, as an active competitor, what are your thoughts on the final results of this competition? Will there be a massive shakeup? </p>",
                      "votes": null,
                      "replies": []
                    }
                  ]
                }
              ]
            }
          ]
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "3072243": "I tried changing every parameter and slighlty different architectures . But can't beat LB - little change and I drop down hard. Is it really that perfect or am I missing something?\nBecause everything has already been tried here?\nSo what else can help?\nSorry for that question",
    "3072258": "This competition feels like a complete game of luck. I’ve done many experiments in the early stages, but they didn’t provide any real advantage. Models with good CV performance are behaving poorly on LB, and even a slight change in the random seed worse alters the results. Everything just feels chaotic and unpredictable. \n\nSo, just wait for the competition to end and see what will Happen",
    "3072469": "> This competition feels like a complete game of luck\n\nThis is indeed a test of luck and lottery @abdmental01",
    "3073760": "> This competition feels like a complete game of luck\n\n> This is indeed a test of luck and lottery\n\nI quit this chaos 2 months ago haha!!\nedit: 3 months ago*",
    "3073916": "I think this is a good decision @eraakash \nThe same thing happened with ICR also, where people with 2-3 submissions performed better than ones with so-called optimization endeavors",
    "3073995": "I can agree that it's a good decision. But I did the stupidest thing I could imagine. When I looked back at my 3mo old code, ... \n\n- I dropped all those columns in train that were not present in test (except target sii) \n\n- I treated that as a regression problem and rounded the predictions to obtain integers\n\n- I was optimizing RMSE.. But there's more - \n\nI tried to \"revamp\" that old approach by all the things I learnt in 3 months but woah, later when I checked the competition code,... The evaluation metric (whose name i never heard before) and basically the objective was entirely different! \n\nThe 'comeback' just backfired so bad for trusting my old self haha!! Maybe aligning my approach with the correct objective can improve my score in the remaining time..? Too late I guess!\n\nBut, as an active competitor, what are your thoughts on the final results of this competition? Will there be a massive shakeup?"
  },
  "source": "meta"
}