{
  "id": 101304,
  "title": "Is any one expecting a Shake up in Private LB",
  "url": "/competitions/generative-dog-images/discussion/101304",
  "author_name": "",
  "post_date": "2019-07-24T18:39:21.696740800Z",
  "votes": 7,
  "comment_count": 11,
  "views": 0,
  "content": "<p>With the new set of rules specific to this competition, is any one expecting a shake up in Private Leaderboard?</p>",
  "messages": [
    {
      "id": "583645",
      "postDate": "07/24/2019 18:39:21",
      "content": "<p>With the new set of rules specific to this competition, is any one expecting a shake up in Private Leaderboard?</p>",
      "rawMarkdown": "With the new set of rules specific to this competition, is any one expecting a shake up in Private Leaderboard?",
      "votes": null
    },
    {
      "id": "583657",
      "postDate": "07/24/2019 18:55:21",
      "content": "<p>These LBs don't mean anything anyway!!</p>",
      "rawMarkdown": "These LBs don't mean anything anyway!!",
      "votes": null
    },
    {
      "id": "585009",
      "postDate": "07/26/2019 19:40:54",
      "content": "<p>It for sure will be interresting to see what the end results will be. If Kaggle filters out all the Memorization GAN's it likely will be a large shake up... reading all the discussion about those it for sure is still a question up to what level that is exceptable and how much effort Kaggle will do to review all kernels.</p>\n\n<p>We will know soon enough ;-)</p>",
      "rawMarkdown": "It for sure will be interresting to see what the end results will be. If Kaggle filters out all the Memorization GAN's it likely will be a large shake up... reading all the discussion about those it for sure is still a question up to what level that is exceptable and how much effort Kaggle will do to review all kernels.\n\nWe will know soon enough ;-)",
      "votes": null
    },
    {
      "id": "585027",
      "postDate": "07/26/2019 20:13:47",
      "content": "<p>Memorizer GANs might just drop down effortlessly. \nI hope the mysterious data and model do wonders and all memorizer GANs score terrible.</p>",
      "rawMarkdown": "Memorizer GANs might just drop down effortlessly. \nI hope the mysterious data and model do wonders and all memorizer GANs score terrible.",
      "votes": null
    },
    {
      "id": "585030",
      "postDate": "07/26/2019 20:15:55",
      "content": "<p>How about DCGANS?</p>",
      "rawMarkdown": "How about DCGANS?",
      "votes": null
    },
    {
      "id": "585048",
      "postDate": "07/26/2019 21:00:03",
      "content": "<p>My interpretation is that as long as you have a generator and discriminator working in the right way it should likely be OK. I'am not sure if it is then relevant whether you are using DCGAN, WGAN, GAN or any of the other variants.</p>\n\n<p>The question is...where does memorization stop and generalization start? To be good at 'generalizing' or creating dog images you need to somehow memorize that they have a head, tail, 4 legs, some kind of collored fur etc.</p>",
      "rawMarkdown": "My interpretation is that as long as you have a generator and discriminator working in the right way it should likely be OK. I'am not sure if it is then relevant whether you are using DCGAN, WGAN, GAN or any of the other variants.\n\nThe question is...where does memorization stop and generalization start? To be good at 'generalizing' or creating dog images you need to somehow memorize that they have a head, tail, 4 legs, some kind of collored fur etc.",
      "votes": null
    },
    {
      "id": "585132",
      "postDate": "07/27/2019 02:18:19",
      "content": "<p>Ahah, I”d rather let someone with 20+ lb score win, I'll be disappointed if kaggle let the 7.2x crowd win</p>",
      "rawMarkdown": "Ahah, I”d rather let someone with 20+ lb score win, I'll be disappointed if kaggle let the 7.2x crowd win",
      "votes": null
    },
    {
      "id": "586459",
      "postDate": "07/29/2019 07:41:27",
      "content": "<p>Could you please tell me where can I find the 'new set of rules', thank you.</p>",
      "rawMarkdown": "Could you please tell me where can I find the 'new set of rules', thank you.",
      "votes": null
    },
    {
      "id": "587999",
      "postDate": "07/30/2019 02:31:42",
      "content": "<p><a href=\"/rsmits\">@rsmits</a>: you are right! I do hope that the host also make some consideration on this paper: Towards GAN Benchmarks which require Generalization: <a href=\"https://openreview.net/pdf?id=HkxKH2AcFm\">https://openreview.net/pdf?id=HkxKH2AcFm</a></p>",
      "rawMarkdown": "rsmits: you are right! I do hope that the host also make some consideration on this paper: Towards GAN Benchmarks which require Generalization: https://openreview.net/pdf?id=HkxKH2AcFm",
      "votes": null
    },
    {
      "id": "588010",
      "postDate": "07/30/2019 02:56:13",
      "content": "<p>You may want to check out Wendy Kan's replies in the Rules clarification on Generative Models - </p>\n\n<p><a href=\"https://www.kaggle.com/c/generative-dog-images/discussion/98183#572299\">https://www.kaggle.com/c/generative-dog-images/discussion/98183#572299</a></p>\n\n<p><a href=\"https://www.kaggle.com/c/generative-dog-images/discussion/98183#573181\">https://www.kaggle.com/c/generative-dog-images/discussion/98183#573181</a></p>",
      "rawMarkdown": "You may want to check out Wendy Kan's replies in the Rules clarification on Generative Models - \n\nhttps://www.kaggle.com/c/generative-dog-images/discussion/98183#572299\n\nhttps://www.kaggle.com/c/generative-dog-images/discussion/98183#573181",
      "votes": null
    },
    {
      "id": "588013",
      "postDate": "07/30/2019 03:01:14",
      "content": "<p>Thanks, that's what I want.</p>",
      "rawMarkdown": "Thanks, that's what I want.",
      "votes": null
    },
    {
      "id": "588247",
      "postDate": "07/30/2019 10:34:03",
      "content": "<blockquote>\n  <p>Is any one expecting a Shake up in Private LB</p>\n</blockquote>\n\n<p>A big one. Biggest of all times because LB makes no sense for this competition. This competition is more of a trial version for GAN based upcoming competitions, if this is successful.</p>",
      "rawMarkdown": "&gt; Is any one expecting a Shake up in Private LB\n\nA big one. Biggest of all times because LB makes no sense for this competition. This competition is more of a trial version for GAN based upcoming competitions, if this is successful.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 583657,
      "author_name": "francoisdubois",
      "author_url": "",
      "post_date": "07/24/2019 18:55:21",
      "content": "<p>These LBs don't mean anything anyway!!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 585009,
      "author_name": "rsmits",
      "author_url": "",
      "post_date": "07/26/2019 19:40:54",
      "content": "<p>It for sure will be interresting to see what the end results will be. If Kaggle filters out all the Memorization GAN's it likely will be a large shake up... reading all the discussion about those it for sure is still a question up to what level that is exceptable and how much effort Kaggle will do to review all kernels.</p>\n\n<p>We will know soon enough ;-)</p>",
      "votes": null,
      "replies": [
        {
          "id": 585027,
          "author_name": "timetraveller98",
          "author_url": "",
          "post_date": "07/26/2019 20:13:47",
          "content": "<p>Memorizer GANs might just drop down effortlessly. \nI hope the mysterious data and model do wonders and all memorizer GANs score terrible.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 585030,
          "author_name": "roydatascience",
          "author_url": "",
          "post_date": "07/26/2019 20:15:55",
          "content": "<p>How about DCGANS?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 585048,
          "author_name": "rsmits",
          "author_url": "",
          "post_date": "07/26/2019 21:00:03",
          "content": "<p>My interpretation is that as long as you have a generator and discriminator working in the right way it should likely be OK. I'am not sure if it is then relevant whether you are using DCGAN, WGAN, GAN or any of the other variants.</p>\n\n<p>The question is...where does memorization stop and generalization start? To be good at 'generalizing' or creating dog images you need to somehow memorize that they have a head, tail, 4 legs, some kind of collored fur etc.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 587999,
          "author_name": "projdev",
          "author_url": "",
          "post_date": "07/30/2019 02:31:42",
          "content": "<p><a href=\"/rsmits\">@rsmits</a>: you are right! I do hope that the host also make some consideration on this paper: Towards GAN Benchmarks which require Generalization: <a href=\"https://openreview.net/pdf?id=HkxKH2AcFm\">https://openreview.net/pdf?id=HkxKH2AcFm</a></p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 585132,
      "author_name": "meistermorxrc",
      "author_url": "",
      "post_date": "07/27/2019 02:18:19",
      "content": "<p>Ahah, I”d rather let someone with 20+ lb score win, I'll be disappointed if kaggle let the 7.2x crowd win</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 586459,
      "author_name": "leonshangguan",
      "author_url": "",
      "post_date": "07/29/2019 07:41:27",
      "content": "<p>Could you please tell me where can I find the 'new set of rules', thank you.</p>",
      "votes": null,
      "replies": [
        {
          "id": 588010,
          "author_name": "something4kag",
          "author_url": "",
          "post_date": "07/30/2019 02:56:13",
          "content": "<p>You may want to check out Wendy Kan's replies in the Rules clarification on Generative Models - </p>\n\n<p><a href=\"https://www.kaggle.com/c/generative-dog-images/discussion/98183#572299\">https://www.kaggle.com/c/generative-dog-images/discussion/98183#572299</a></p>\n\n<p><a href=\"https://www.kaggle.com/c/generative-dog-images/discussion/98183#573181\">https://www.kaggle.com/c/generative-dog-images/discussion/98183#573181</a></p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 588013,
          "author_name": "leonshangguan",
          "author_url": "",
          "post_date": "07/30/2019 03:01:14",
          "content": "<p>Thanks, that's what I want.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 588247,
      "author_name": "cyberia",
      "author_url": "",
      "post_date": "07/30/2019 10:34:03",
      "content": "<blockquote>\n  <p>Is any one expecting a Shake up in Private LB</p>\n</blockquote>\n\n<p>A big one. Biggest of all times because LB makes no sense for this competition. This competition is more of a trial version for GAN based upcoming competitions, if this is successful.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "583645": "With the new set of rules specific to this competition, is any one expecting a shake up in Private Leaderboard?",
    "583657": "These LBs don't mean anything anyway!!",
    "585009": "It for sure will be interresting to see what the end results will be. If Kaggle filters out all the Memorization GAN's it likely will be a large shake up... reading all the discussion about those it for sure is still a question up to what level that is exceptable and how much effort Kaggle will do to review all kernels.\n\nWe will know soon enough ;-)",
    "585027": "Memorizer GANs might just drop down effortlessly. \nI hope the mysterious data and model do wonders and all memorizer GANs score terrible.",
    "585030": "How about DCGANS?",
    "585048": "My interpretation is that as long as you have a generator and discriminator working in the right way it should likely be OK. I'am not sure if it is then relevant whether you are using DCGAN, WGAN, GAN or any of the other variants.\n\nThe question is...where does memorization stop and generalization start? To be good at 'generalizing' or creating dog images you need to somehow memorize that they have a head, tail, 4 legs, some kind of collored fur etc.",
    "585132": "Ahah, I”d rather let someone with 20+ lb score win, I'll be disappointed if kaggle let the 7.2x crowd win",
    "586459": "Could you please tell me where can I find the 'new set of rules', thank you.",
    "587999": "rsmits: you are right! I do hope that the host also make some consideration on this paper: Towards GAN Benchmarks which require Generalization: https://openreview.net/pdf?id=HkxKH2AcFm",
    "588010": "You may want to check out Wendy Kan's replies in the Rules clarification on Generative Models - \n\nhttps://www.kaggle.com/c/generative-dog-images/discussion/98183#572299\n\nhttps://www.kaggle.com/c/generative-dog-images/discussion/98183#573181",
    "588013": "Thanks, that's what I want.",
    "588247": "&gt; Is any one expecting a Shake up in Private LB\n\nA big one. Biggest of all times because LB makes no sense for this competition. This competition is more of a trial version for GAN based upcoming competitions, if this is successful."
  },
  "source": "meta"
}