{
  "id": 18599,
  "title": "Questions about Hand-labeling",
  "url": "/competitions/second-annual-data-science-bowl/discussion/18599",
  "author_name": "",
  "post_date": "2016-01-27T13:52:27.240Z",
  "votes": null,
  "comment_count": 9,
  "views": 1638,
  "content": "<p>From the rules:</p>\n\n<blockquote>\n  <p>Hand-labeling is allowed on the training dataset only. Hand-labeling\n  is not permitted on test data and will be grounds for\n  disqualification.</p>\n</blockquote>\n\n<p>1) Does this mean we can also hand-label the images in the &quot;validate&quot; folder?</p>\n\n<p>2) Should these methods/information be included when uploading the model? If so, how exactly?</p>",
  "messages": [
    {
      "id": "105931",
      "postDate": "01/27/2016 13:52:27",
      "content": "<p>From the rules:</p>\n\n<blockquote>\n  <p>Hand-labeling is allowed on the training dataset only. Hand-labeling\n  is not permitted on test data and will be grounds for\n  disqualification.</p>\n</blockquote>\n\n<p>1) Does this mean we can also hand-label the images in the &quot;validate&quot; folder?</p>\n\n<p>2) Should these methods/information be included when uploading the model? If so, how exactly?</p>",
      "rawMarkdown": "From the rules:\r\n> Hand-labeling is allowed on the training dataset only. Hand-labeling\r\n> is not permitted on test data and will be grounds for\r\n> disqualification.\r\n\r\n1) Does this mean we can also hand-label the images in the \"validate\" folder?\r\n\r\n2) Should these methods/information be included when uploading the model? If so, how exactly?",
      "votes": null
    },
    {
      "id": "105933",
      "postDate": "01/27/2016 14:07:10",
      "content": "<ul>\n<li>During phase one, you can hand label the training set only.</li>\n<li>During phase two, after we have released the validation set labels, you can hand label the training and validation set only.</li>\n</ul>\n\n<p>Yes, you should include brief instructions/explanations in the model's README when you upload it.</p>",
      "rawMarkdown": "During phase one, you can hand label the training set only.\r\n - During phase two, after we have released the validation set labels, you can hand label the training and validation set only.\r\n\r\nYes, you should include brief instructions/explanations in the model's README when you upload it.",
      "votes": null
    },
    {
      "id": "105937",
      "postDate": "01/27/2016 14:29:39",
      "content": "<p>Thanks, sounds good.</p>",
      "rawMarkdown": "Thanks, sounds good.",
      "votes": null
    },
    {
      "id": "105989",
      "postDate": "01/27/2016 20:26:37",
      "content": "<p>Actually, I have another question. Is there an example model submission somewhere that can be used as a template to follow? Even if the example is not specific to this competition it would be helpful to see what kind of directory layout, level of detail in the readme/comments, etc is expected.</p>",
      "rawMarkdown": "Actually, I have another question. Is there an example model submission somewhere that can be used as a template to follow? Even if the example is not specific to this competition it would be helpful to see what kind of directory layout, level of detail in the readme/comments, etc is expected.",
      "votes": null
    },
    {
      "id": "105992",
      "postDate": "01/27/2016 20:38:52",
      "content": "<p>Check out <a href=\"https://www.kaggle.com/wiki/ModelSubmissionBestPractices\">https://www.kaggle.com/wiki/ModelSubmissionBestPractices</a>. We keep the process informal since models come in a lot of shapes and sizes. The goal of the stage 2 model upload is reproducibility/accountability, not necessarily to write lengthy report.</p>\n\n<p>If you win, you'll receive a template and be asked to write up a more thorough report.</p>",
      "rawMarkdown": "Check out https://www.kaggle.com/wiki/ModelSubmissionBestPractices. We keep the process informal since models come in a lot of shapes and sizes. The goal of the stage 2 model upload is reproducibility/accountability, not necessarily to write lengthy report.\r\n\r\nIf you win, you'll receive a template and be asked to write up a more thorough report.",
      "votes": null
    },
    {
      "id": "108654",
      "postDate": "02/18/2016 23:14:44",
      "content": "<p>@William,</p>\n\n<p>Newbie (and late) here! Is it possible to change the code between March 7th and March 14th?</p>\n\n<p>Thanks in advance,</p>\n\n<p>Tommy</p>",
      "rawMarkdown": "William,\r\n\r\nNewbie (and late) here! Is it possible to change the code between March 7th and March 14th?\r\n\r\nThanks in advance,\r\n\r\nTommy",
      "votes": null
    },
    {
      "id": "108668",
      "postDate": "02/19/2016 02:17:01",
      "content": "<p>Hi Tommy,</p>\n\n<p>Welcome! It is not permitted to change your model after March 7th.</p>",
      "rawMarkdown": "Hi Tommy,\r\n\r\nWelcome! It is not permitted to change your model after March 7th.",
      "votes": null
    },
    {
      "id": "109809",
      "postDate": "03/01/2016 02:36:31",
      "content": "<p><strong>Kaggle Admin</strong> William Cukierski, i have some questions about what means hand-labeling.</p>\n\n<p>Can you help me ?</p>\n\n<p>If i apply a mathematical function in <strong>all images</strong> to find some pattern it's  hand-labeling ? </p>\n\n<p>And if I apply a logical function that search a logical pattern in <strong>all images and reacts uniquely for a particular pattern that can exists in one or more images</strong> it's hand-labeling ?</p>\n\n<p>The question is, not always neural networks solve problems and we have to apply mathematical transformations or even logical decisions in the process.</p>",
      "rawMarkdown": "**Kaggle Admin** William Cukierski, i have some questions about what means hand-labeling.\r\n\r\nCan you help me ?\r\n\r\nIf i apply a mathematical function in **all images** to find some pattern it's  hand-labeling ? \r\n\r\nAnd if I apply a logical function that search a logical pattern in **all images and reacts uniquely for a particular pattern that can exists in one or more images** it's hand-labeling ?\r\n\r\nThe question is, not always neural networks solve problems and we have to apply mathematical transformations or even logical decisions in the process.",
      "votes": null
    },
    {
      "id": "109910",
      "postDate": "03/01/2016 13:48:10",
      "content": "<p>Hi Alvaro,</p>\n\n<p>Applying a mathematical function is not hand labeling. Hand labeling is any process (circling a ventricle, picking certain images to ignore in training, etc.) which you have to manually redo each time you encounter new data.</p>",
      "rawMarkdown": "Hi Alvaro,\r\n\r\nApplying a mathematical function is not hand labeling. Hand labeling is any process (circling a ventricle, picking certain images to ignore in training, etc.) which you have to manually redo each time you encounter new data.",
      "votes": null
    },
    {
      "id": "109929",
      "postDate": "03/01/2016 15:17:14",
      "content": "<p>Thanks very much William. :D</p>",
      "rawMarkdown": "Thanks very much William. :D",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 105933,
      "author_name": "wcukierski",
      "author_url": "",
      "post_date": "01/27/2016 14:07:10",
      "content": "<ul>\n<li>During phase one, you can hand label the training set only.</li>\n<li>During phase two, after we have released the validation set labels, you can hand label the training and validation set only.</li>\n</ul>\n\n<p>Yes, you should include brief instructions/explanations in the model's README when you upload it.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 105937,
      "author_name": "halfwit",
      "author_url": "",
      "post_date": "01/27/2016 14:29:39",
      "content": "<p>Thanks, sounds good.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 105989,
      "author_name": "halfwit",
      "author_url": "",
      "post_date": "01/27/2016 20:26:37",
      "content": "<p>Actually, I have another question. Is there an example model submission somewhere that can be used as a template to follow? Even if the example is not specific to this competition it would be helpful to see what kind of directory layout, level of detail in the readme/comments, etc is expected.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 105992,
      "author_name": "wcukierski",
      "author_url": "",
      "post_date": "01/27/2016 20:38:52",
      "content": "<p>Check out <a href=\"https://www.kaggle.com/wiki/ModelSubmissionBestPractices\">https://www.kaggle.com/wiki/ModelSubmissionBestPractices</a>. We keep the process informal since models come in a lot of shapes and sizes. The goal of the stage 2 model upload is reproducibility/accountability, not necessarily to write lengthy report.</p>\n\n<p>If you win, you'll receive a template and be asked to write up a more thorough report.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 108654,
      "author_name": "tommyb",
      "author_url": "",
      "post_date": "02/18/2016 23:14:44",
      "content": "<p>@William,</p>\n\n<p>Newbie (and late) here! Is it possible to change the code between March 7th and March 14th?</p>\n\n<p>Thanks in advance,</p>\n\n<p>Tommy</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 108668,
      "author_name": "wcukierski",
      "author_url": "",
      "post_date": "02/19/2016 02:17:01",
      "content": "<p>Hi Tommy,</p>\n\n<p>Welcome! It is not permitted to change your model after March 7th.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 109809,
      "author_name": "alvaroosvaldo",
      "author_url": "",
      "post_date": "03/01/2016 02:36:31",
      "content": "<p><strong>Kaggle Admin</strong> William Cukierski, i have some questions about what means hand-labeling.</p>\n\n<p>Can you help me ?</p>\n\n<p>If i apply a mathematical function in <strong>all images</strong> to find some pattern it's  hand-labeling ? </p>\n\n<p>And if I apply a logical function that search a logical pattern in <strong>all images and reacts uniquely for a particular pattern that can exists in one or more images</strong> it's hand-labeling ?</p>\n\n<p>The question is, not always neural networks solve problems and we have to apply mathematical transformations or even logical decisions in the process.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 109910,
      "author_name": "wcukierski",
      "author_url": "",
      "post_date": "03/01/2016 13:48:10",
      "content": "<p>Hi Alvaro,</p>\n\n<p>Applying a mathematical function is not hand labeling. Hand labeling is any process (circling a ventricle, picking certain images to ignore in training, etc.) which you have to manually redo each time you encounter new data.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 109929,
      "author_name": "alvaroosvaldo",
      "author_url": "",
      "post_date": "03/01/2016 15:17:14",
      "content": "<p>Thanks very much William. :D</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "105931": "From the rules:\r\n> Hand-labeling is allowed on the training dataset only. Hand-labeling\r\n> is not permitted on test data and will be grounds for\r\n> disqualification.\r\n\r\n1) Does this mean we can also hand-label the images in the \"validate\" folder?\r\n\r\n2) Should these methods/information be included when uploading the model? If so, how exactly?",
    "105933": "During phase one, you can hand label the training set only.\r\n - During phase two, after we have released the validation set labels, you can hand label the training and validation set only.\r\n\r\nYes, you should include brief instructions/explanations in the model's README when you upload it.",
    "105937": "Thanks, sounds good.",
    "105989": "Actually, I have another question. Is there an example model submission somewhere that can be used as a template to follow? Even if the example is not specific to this competition it would be helpful to see what kind of directory layout, level of detail in the readme/comments, etc is expected.",
    "105992": "Check out https://www.kaggle.com/wiki/ModelSubmissionBestPractices. We keep the process informal since models come in a lot of shapes and sizes. The goal of the stage 2 model upload is reproducibility/accountability, not necessarily to write lengthy report.\r\n\r\nIf you win, you'll receive a template and be asked to write up a more thorough report.",
    "108654": "William,\r\n\r\nNewbie (and late) here! Is it possible to change the code between March 7th and March 14th?\r\n\r\nThanks in advance,\r\n\r\nTommy",
    "108668": "Hi Tommy,\r\n\r\nWelcome! It is not permitted to change your model after March 7th.",
    "109809": "**Kaggle Admin** William Cukierski, i have some questions about what means hand-labeling.\r\n\r\nCan you help me ?\r\n\r\nIf i apply a mathematical function in **all images** to find some pattern it's  hand-labeling ? \r\n\r\nAnd if I apply a logical function that search a logical pattern in **all images and reacts uniquely for a particular pattern that can exists in one or more images** it's hand-labeling ?\r\n\r\nThe question is, not always neural networks solve problems and we have to apply mathematical transformations or even logical decisions in the process.",
    "109910": "Hi Alvaro,\r\n\r\nApplying a mathematical function is not hand labeling. Hand labeling is any process (circling a ventricle, picking certain images to ignore in training, etc.) which you have to manually redo each time you encounter new data.",
    "109929": "Thanks very much William. :D"
  },
  "source": "meta"
}