{
  "id": 78667,
  "title": "Limit of 2 submissions per day",
  "url": "/competitions/vsb-power-line-fault-detection/discussion/78667",
  "author_name": "",
  "post_date": "2019-01-26T18:16:34.310725300Z",
  "votes": 1,
  "comment_count": 4,
  "views": 0,
  "content": "<p>Is there an special reason why this competition has a limit of 2 submissions per day?</p>\n\n<p>Most of the past competitions here on Kaggle, and actually all of the other, currently, featured competitions have a 5 submissions per day limit.</p>\n\n<p>It seems a bit rough to have a 2 submissions limit. I'm having a hard time coming up on where to improve on my model/pre-processing efforts to improve the score since the results on LB seems quite random with small changes on the process.</p>\n\n<p>I'm guessing a lot of people are using Keras for Neural Networks and having a hard time to come up with reproducible results, I've been using PyTorch and all of my scores are reproducible and yet I can't seem to find I path that leads to better results. </p>\n\n<p>I know I should trust my CV scores more than the LB scores, but a higher limit would allow everyone to experiment more and have a better idea on what improves our models. Any thoughts?</p>",
  "messages": [
    {
      "id": "461688",
      "postDate": "01/26/2019 18:16:34",
      "content": "<p>Is there an special reason why this competition has a limit of 2 submissions per day?</p>\n\n<p>Most of the past competitions here on Kaggle, and actually all of the other, currently, featured competitions have a 5 submissions per day limit.</p>\n\n<p>It seems a bit rough to have a 2 submissions limit. I'm having a hard time coming up on where to improve on my model/pre-processing efforts to improve the score since the results on LB seems quite random with small changes on the process.</p>\n\n<p>I'm guessing a lot of people are using Keras for Neural Networks and having a hard time to come up with reproducible results, I've been using PyTorch and all of my scores are reproducible and yet I can't seem to find I path that leads to better results. </p>\n\n<p>I know I should trust my CV scores more than the LB scores, but a higher limit would allow everyone to experiment more and have a better idea on what improves our models. Any thoughts?</p>",
      "rawMarkdown": "Is there an special reason why this competition has a limit of 2 submissions per day?\n\nMost of the past competitions here on Kaggle, and actually all of the other, currently, featured competitions have a 5 submissions per day limit.\n\nIt seems a bit rough to have a 2 submissions limit. I'm having a hard time coming up on where to improve on my model/pre-processing efforts to improve the score since the results on LB seems quite random with small changes on the process.\n\nI'm guessing a lot of people are using Keras for Neural Networks and having a hard time to come up with reproducible results, I've been using PyTorch and all of my scores are reproducible and yet I can't seem to find I path that leads to better results. \n\nI know I should trust my CV scores more than the LB scores, but a higher limit would allow everyone to experiment more and have a better idea on what improves our models. Any thoughts?",
      "votes": null
    },
    {
      "id": "461820",
      "postDate": "01/27/2019 04:08:55",
      "content": "<p>This competition uses 57% of data for public testing, meaning the LB may be even more reliable than CV. I guess by not making life too easy for us they limited the number of submissions.</p>",
      "rawMarkdown": "This competition uses 57% of data for public testing, meaning the LB may be even more reliable than CV. I guess by not making life too easy for us they limited the number of submissions.",
      "votes": null
    },
    {
      "id": "462179",
      "postDate": "01/27/2019 20:14:50",
      "content": "<p>Once we accept our limits, we go beyond them. </p>\n\n<p>Albert Einstein</p>",
      "rawMarkdown": "Once we accept our limits, we go beyond them. \n\nAlbert Einstein",
      "votes": null
    },
    {
      "id": "462740",
      "postDate": "01/28/2019 20:10:53",
      "content": "<p>2 submission per day is to avoid probing, I guess.</p>",
      "rawMarkdown": "2 submission per day is to avoid probing, I guess.",
      "votes": null
    },
    {
      "id": "463173",
      "postDate": "01/29/2019 15:13:52",
      "content": "<p>I agree 2 submission per day is much restrictive. Can it pushed up as time progresses?</p>",
      "rawMarkdown": "I agree 2 submission per day is much restrictive. Can it pushed up as time progresses?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 461820,
      "author_name": "suicaokhoailang",
      "author_url": "",
      "post_date": "01/27/2019 04:08:55",
      "content": "<p>This competition uses 57% of data for public testing, meaning the LB may be even more reliable than CV. I guess by not making life too easy for us they limited the number of submissions.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 462179,
      "author_name": "a45632",
      "author_url": "",
      "post_date": "01/27/2019 20:14:50",
      "content": "<p>Once we accept our limits, we go beyond them. </p>\n\n<p>Albert Einstein</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 462740,
      "author_name": "adilztn",
      "author_url": "",
      "post_date": "01/28/2019 20:10:53",
      "content": "<p>2 submission per day is to avoid probing, I guess.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 463173,
      "author_name": "padmakartj",
      "author_url": "",
      "post_date": "01/29/2019 15:13:52",
      "content": "<p>I agree 2 submission per day is much restrictive. Can it pushed up as time progresses?</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "461688": "Is there an special reason why this competition has a limit of 2 submissions per day?\n\nMost of the past competitions here on Kaggle, and actually all of the other, currently, featured competitions have a 5 submissions per day limit.\n\nIt seems a bit rough to have a 2 submissions limit. I'm having a hard time coming up on where to improve on my model/pre-processing efforts to improve the score since the results on LB seems quite random with small changes on the process.\n\nI'm guessing a lot of people are using Keras for Neural Networks and having a hard time to come up with reproducible results, I've been using PyTorch and all of my scores are reproducible and yet I can't seem to find I path that leads to better results. \n\nI know I should trust my CV scores more than the LB scores, but a higher limit would allow everyone to experiment more and have a better idea on what improves our models. Any thoughts?",
    "461820": "This competition uses 57% of data for public testing, meaning the LB may be even more reliable than CV. I guess by not making life too easy for us they limited the number of submissions.",
    "462179": "Once we accept our limits, we go beyond them. \n\nAlbert Einstein",
    "462740": "2 submission per day is to avoid probing, I guess.",
    "463173": "I agree 2 submission per day is much restrictive. Can it pushed up as time progresses?"
  },
  "source": "meta"
}