{
  "id": 15102,
  "title": "Should I enable Scripts on this competition?",
  "url": "/competitions/diabetic-retinopathy-detection/discussion/15102",
  "author_name": "",
  "post_date": "2015-07-07T22:46:29.127Z",
  "votes": 10,
  "comment_count": 8,
  "views": 1746,
  "content": "<p>All - I'm considering enabling <a href=\"kaggle.com/scripts\">Kaggle Scripts</a> on this competition on a tiny subset of the training data.</p>\n\n<p>The current execution environment (CPU only + 10 minute time limit) isn't nearly enough to do anything meaningful on the entire dataset. This means that you wouldn't be able to publish computationally-intensive analytics on this dataset through scripts, or provide sample code that makes submissions through scripts.</p>\n\n<p>If I enabled Scripts on this competition on a small subset of the training data, could you create and publish anything meaningful and interesting based on that? Is there any specific subset of the training data that it would make more sense to use?</p>\n\n<p>If so, I'll go ahead and do it.</p>",
  "messages": [
    {
      "id": "83718",
      "postDate": "07/07/2015 22:46:29",
      "content": "<p>All - I'm considering enabling <a href=\"kaggle.com/scripts\">Kaggle Scripts</a> on this competition on a tiny subset of the training data.</p>\n\n<p>The current execution environment (CPU only + 10 minute time limit) isn't nearly enough to do anything meaningful on the entire dataset. This means that you wouldn't be able to publish computationally-intensive analytics on this dataset through scripts, or provide sample code that makes submissions through scripts.</p>\n\n<p>If I enabled Scripts on this competition on a small subset of the training data, could you create and publish anything meaningful and interesting based on that? Is there any specific subset of the training data that it would make more sense to use?</p>\n\n<p>If so, I'll go ahead and do it.</p>",
      "rawMarkdown": "All - I'm considering enabling [Kaggle Scripts](kaggle.com/scripts) on this competition on a tiny subset of the training data.\r\n\r\nThe current execution environment (CPU only + 10 minute time limit) isn't nearly enough to do anything meaningful on the entire dataset. This means that you wouldn't be able to publish computationally-intensive analytics on this dataset through scripts, or provide sample code that makes submissions through scripts.\r\n\r\nIf I enabled Scripts on this competition on a small subset of the training data, could you create and publish anything meaningful and interesting based on that? Is there any specific subset of the training data that it would make more sense to use?\r\n\r\nIf so, I'll go ahead and do it.",
      "votes": null
    },
    {
      "id": "83725",
      "postDate": "07/07/2015 23:34:14",
      "content": "<p>I'd love a subset of the data to work on. I downloaded part of the archive, but noticed I could not unzip it without the rest. I am curious what the rest of the data set looks like (beyond the small preview zip), if only to visually compare a few &quot;1&quot;'s with &quot;4&quot;'s</p>\n\n<p>Better yet would be to run a image size reduction on the data set. I think Abhishek posted a script once to do that. That would drastically reduce download threshold for those who do not want to (or can) compete with 2MB images.</p>",
      "rawMarkdown": "I'd love a subset of the data to work on. I downloaded part of the archive, but noticed I could not unzip it without the rest. I am curious what the rest of the data set looks like (beyond the small preview zip), if only to visually compare a few \"1\"'s with \"4\"'s\r\n\r\nBetter yet would be to run a image size reduction on the data set. I think Abhishek posted a script once to do that. That would drastically reduce download threshold for those who do not want to (or can) compete with 2MB images.",
      "votes": null
    },
    {
      "id": "83869",
      "postDate": "07/08/2015 22:45:56",
      "content": "<p>We would love to have scripts for this. </p>",
      "rawMarkdown": "We would love to have scripts for this.",
      "votes": null
    },
    {
      "id": "83904",
      "postDate": "07/09/2015 07:22:35",
      "content": "<p>In my humble opinion, I wouldn't enable scripts so late in the competition. In my limited experience (only Otto and West-Nile), having scripts enabled does not necessarily lead to a better competition; in fact quite the opposite with people posting high benchmark codes very late. My score is low in this one, but I have enjoyed (and learned, albeit on my own instead of through scripts and the forum) a lot more here than in my previous two competitions.</p>\n\n<p>Even if it's just a subset and we won't be able to submit through them, the scripts functionality is basically a legitimate and easier way to share hypertuned models.</p>\n\n<p>Would it be possible to enable them <em>after</em> the competition ends, for the folks wanting to try out some things?</p>",
      "rawMarkdown": "In my humble opinion, I wouldn't enable scripts so late in the competition. In my limited experience (only Otto and West-Nile), having scripts enabled does not necessarily lead to a better competition; in fact quite the opposite with people posting high benchmark codes very late. My score is low in this one, but I have enjoyed (and learned, albeit on my own instead of through scripts and the forum) a lot more here than in my previous two competitions.\r\n\r\nEven if it's just a subset and we won't be able to submit through them, the scripts functionality is basically a legitimate and easier way to share hypertuned models.\r\n\r\nWould it be possible to enable them _after_ the competition ends, for the folks wanting to try out some things?",
      "votes": null
    },
    {
      "id": "83947",
      "postDate": "07/09/2015 15:10:02",
      "content": "<p>I think it's better to have the scripts where we can run something... Especially not everyone has supercomputer and download the data is quite painful. It would be better for everyone to try something. Reduce the image size would also be helpful since learning is much more important than the final rank....</p>",
      "rawMarkdown": "I think it's better to have the scripts where we can run something... Especially not everyone has supercomputer and download the data is quite painful. It would be better for everyone to try something. Reduce the image size would also be helpful since learning is much more important than the final rank....",
      "votes": null
    },
    {
      "id": "83966",
      "postDate": "07/09/2015 19:55:54",
      "content": "<p>+1 for a more usable sample data set (what @Triskelion said)\n+1 for it being too late to enable scripts, but seeing them after the end would be nice (@Timothy Man)</p>\n\n<p>I would loved to have seen them from the start; these image manipulation ones are interesting to me, but the data set is a bit large to even get downloaded, so I haven't tried yet... maybe this weekend. :-)</p>",
      "rawMarkdown": "1 for a more usable sample data set (what @Triskelion said)\r\n+1 for it being too late to enable scripts, but seeing them after the end would be nice (@Timothy Man)\r\n\r\nI would loved to have seen them from the start; these image manipulation ones are interesting to me, but the data set is a bit large to even get downloaded, so I haven't tried yet... maybe this weekend. :-)",
      "votes": null
    },
    {
      "id": "83971",
      "postDate": "07/09/2015 20:08:03",
      "content": "<p>@ChipMonkey \nThe same here, there're few competitions like this but haven't get a chance to unzip the data yet....</p>",
      "rawMarkdown": "ChipMonkey \r\nThe same here, there're few competitions like this but haven't get a chance to unzip the data yet....",
      "votes": null
    },
    {
      "id": "84162",
      "postDate": "07/11/2015 20:50:24",
      "content": "<p>I would really love if scripts were enabled. I do not have that much experience with image processing, so I would like to use knowledge that other users have to increase my understanding of the subject.</p>\n\n<p>For example, right now I have problems with cropping. Images are slightly different so my python script or bash command for imagemagic works on the some of them and does not on others. </p>",
      "rawMarkdown": "I would really love if scripts were enabled. I do not have that much experience with image processing, so I would like to use knowledge that other users have to increase my understanding of the subject.\r\n\r\nFor example, right now I have problems with cropping. Images are slightly different so my python script or bash command for imagemagic works on the some of them and does not on others.",
      "votes": null
    },
    {
      "id": "84303",
      "postDate": "07/13/2015 19:13:25",
      "content": "<p>Thanks for all the input!</p>\n\n<p>We've decided to enable scripts in a limited way (on a small ~1gb sample of the training dataset), but to hold off until the end of the competition to do so.</p>",
      "rawMarkdown": "Thanks for all the input!\r\n\r\nWe've decided to enable scripts in a limited way (on a small ~1gb sample of the training dataset), but to hold off until the end of the competition to do so.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 83725,
      "author_name": "triskelion",
      "author_url": "",
      "post_date": "07/07/2015 23:34:14",
      "content": "<p>I'd love a subset of the data to work on. I downloaded part of the archive, but noticed I could not unzip it without the rest. I am curious what the rest of the data set looks like (beyond the small preview zip), if only to visually compare a few &quot;1&quot;'s with &quot;4&quot;'s</p>\n\n<p>Better yet would be to run a image size reduction on the data set. I think Abhishek posted a script once to do that. That would drastically reduce download threshold for those who do not want to (or can) compete with 2MB images.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 83869,
      "author_name": "duttaroy",
      "author_url": "",
      "post_date": "07/08/2015 22:45:56",
      "content": "<p>We would love to have scripts for this. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 83904,
      "author_name": "timothyman",
      "author_url": "",
      "post_date": "07/09/2015 07:22:35",
      "content": "<p>In my humble opinion, I wouldn't enable scripts so late in the competition. In my limited experience (only Otto and West-Nile), having scripts enabled does not necessarily lead to a better competition; in fact quite the opposite with people posting high benchmark codes very late. My score is low in this one, but I have enjoyed (and learned, albeit on my own instead of through scripts and the forum) a lot more here than in my previous two competitions.</p>\n\n<p>Even if it's just a subset and we won't be able to submit through them, the scripts functionality is basically a legitimate and easier way to share hypertuned models.</p>\n\n<p>Would it be possible to enable them <em>after</em> the competition ends, for the folks wanting to try out some things?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 83947,
      "author_name": "lancerts",
      "author_url": "",
      "post_date": "07/09/2015 15:10:02",
      "content": "<p>I think it's better to have the scripts where we can run something... Especially not everyone has supercomputer and download the data is quite painful. It would be better for everyone to try something. Reduce the image size would also be helpful since learning is much more important than the final rank....</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 83966,
      "author_name": "chipmonkey",
      "author_url": "",
      "post_date": "07/09/2015 19:55:54",
      "content": "<p>+1 for a more usable sample data set (what @Triskelion said)\n+1 for it being too late to enable scripts, but seeing them after the end would be nice (@Timothy Man)</p>\n\n<p>I would loved to have seen them from the start; these image manipulation ones are interesting to me, but the data set is a bit large to even get downloaded, so I haven't tried yet... maybe this weekend. :-)</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 83971,
      "author_name": "lancerts",
      "author_url": "",
      "post_date": "07/09/2015 20:08:03",
      "content": "<p>@ChipMonkey \nThe same here, there're few competitions like this but haven't get a chance to unzip the data yet....</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 84162,
      "author_name": "iglovikov",
      "author_url": "",
      "post_date": "07/11/2015 20:50:24",
      "content": "<p>I would really love if scripts were enabled. I do not have that much experience with image processing, so I would like to use knowledge that other users have to increase my understanding of the subject.</p>\n\n<p>For example, right now I have problems with cropping. Images are slightly different so my python script or bash command for imagemagic works on the some of them and does not on others. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 84303,
      "author_name": "benhamner",
      "author_url": "",
      "post_date": "07/13/2015 19:13:25",
      "content": "<p>Thanks for all the input!</p>\n\n<p>We've decided to enable scripts in a limited way (on a small ~1gb sample of the training dataset), but to hold off until the end of the competition to do so.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "83718": "All - I'm considering enabling [Kaggle Scripts](kaggle.com/scripts) on this competition on a tiny subset of the training data.\r\n\r\nThe current execution environment (CPU only + 10 minute time limit) isn't nearly enough to do anything meaningful on the entire dataset. This means that you wouldn't be able to publish computationally-intensive analytics on this dataset through scripts, or provide sample code that makes submissions through scripts.\r\n\r\nIf I enabled Scripts on this competition on a small subset of the training data, could you create and publish anything meaningful and interesting based on that? Is there any specific subset of the training data that it would make more sense to use?\r\n\r\nIf so, I'll go ahead and do it.",
    "83725": "I'd love a subset of the data to work on. I downloaded part of the archive, but noticed I could not unzip it without the rest. I am curious what the rest of the data set looks like (beyond the small preview zip), if only to visually compare a few \"1\"'s with \"4\"'s\r\n\r\nBetter yet would be to run a image size reduction on the data set. I think Abhishek posted a script once to do that. That would drastically reduce download threshold for those who do not want to (or can) compete with 2MB images.",
    "83869": "We would love to have scripts for this.",
    "83904": "In my humble opinion, I wouldn't enable scripts so late in the competition. In my limited experience (only Otto and West-Nile), having scripts enabled does not necessarily lead to a better competition; in fact quite the opposite with people posting high benchmark codes very late. My score is low in this one, but I have enjoyed (and learned, albeit on my own instead of through scripts and the forum) a lot more here than in my previous two competitions.\r\n\r\nEven if it's just a subset and we won't be able to submit through them, the scripts functionality is basically a legitimate and easier way to share hypertuned models.\r\n\r\nWould it be possible to enable them _after_ the competition ends, for the folks wanting to try out some things?",
    "83947": "I think it's better to have the scripts where we can run something... Especially not everyone has supercomputer and download the data is quite painful. It would be better for everyone to try something. Reduce the image size would also be helpful since learning is much more important than the final rank....",
    "83966": "1 for a more usable sample data set (what @Triskelion said)\r\n+1 for it being too late to enable scripts, but seeing them after the end would be nice (@Timothy Man)\r\n\r\nI would loved to have seen them from the start; these image manipulation ones are interesting to me, but the data set is a bit large to even get downloaded, so I haven't tried yet... maybe this weekend. :-)",
    "83971": "ChipMonkey \r\nThe same here, there're few competitions like this but haven't get a chance to unzip the data yet....",
    "84162": "I would really love if scripts were enabled. I do not have that much experience with image processing, so I would like to use knowledge that other users have to increase my understanding of the subject.\r\n\r\nFor example, right now I have problems with cropping. Images are slightly different so my python script or bash command for imagemagic works on the some of them and does not on others.",
    "84303": "Thanks for all the input!\r\n\r\nWe've decided to enable scripts in a limited way (on a small ~1gb sample of the training dataset), but to hold off until the end of the competition to do so."
  },
  "source": "meta"
}