{
  "id": 104915,
  "title": "Improvements after 0.8 LB",
  "url": "/competitions/aptos2019-blindness-detection/discussion/104915",
  "author_name": "",
  "post_date": "2019-08-20T02:16:24.740624500Z",
  "votes": 9,
  "comment_count": 12,
  "views": 0,
  "content": "<p>Im finding it difficult to improve my LB score after reaching 0.8. Can you guys share any tips/tricks/hints that helped you improve beyond 0.8</p>",
  "messages": [
    {
      "id": "603193",
      "postDate": "08/20/2019 02:16:24",
      "content": "<p>Im finding it difficult to improve my LB score after reaching 0.8. Can you guys share any tips/tricks/hints that helped you improve beyond 0.8</p>",
      "rawMarkdown": "Im finding it difficult to improve my LB score after reaching 0.8. Can you guys share any tips/tricks/hints that helped you improve beyond 0.8",
      "votes": null
    },
    {
      "id": "603227",
      "postDate": "08/20/2019 03:15:10",
      "content": "<p>just do ensemble many and many models =)))</p>",
      "rawMarkdown": "just do ensemble many and many models =)))",
      "votes": null
    },
    {
      "id": "604581",
      "postDate": "08/21/2019 14:53:47",
      "content": "<p>Improvements after 0.8? I'm even struggling to hit that mark. Would be nice if you guys can share any tips/tricks/hints. LOL!</p>",
      "rawMarkdown": "Improvements after 0.8? I'm even struggling to hit that mark. Would be nice if you guys can share any tips/tricks/hints. LOL!",
      "votes": null
    },
    {
      "id": "604725",
      "postDate": "08/21/2019 17:37:18",
      "content": "<p>Following the suggestions given by DrHB and Risabh should give you 0.8</p>",
      "rawMarkdown": "Following the suggestions given by DrHB and Risabh should give you 0.8",
      "votes": null
    },
    {
      "id": "605023",
      "postDate": "08/22/2019 01:16:18",
      "content": "<p>What method can arrive 0.8, I now 0.78？</p>",
      "rawMarkdown": "What method can arrive 0.8, I now 0.78？",
      "votes": null
    },
    {
      "id": "605119",
      "postDate": "08/22/2019 04:48:45",
      "content": "<p>I found DrHB's thread but Couldn't find Risabh. Could you link it here?</p>",
      "rawMarkdown": "I found DrHB's thread but Couldn't find Risabh. Could you link it here?",
      "votes": null
    },
    {
      "id": "605301",
      "postDate": "08/22/2019 08:35:03",
      "content": "<p>Look for his comments in <a href=\"https://www.kaggle.com/c/aptos2019-blindness-detection/discussion/97860\">https://www.kaggle.com/c/aptos2019-blindness-detection/discussion/97860</a>, <a href=\"https://www.kaggle.com/c/aptos2019-blindness-detection/discussion/100815\">https://www.kaggle.com/c/aptos2019-blindness-detection/discussion/100815</a> &amp; in <a href=\"https://www.kaggle.com/c/aptos2019-blindness-detection/discussion/99603\">https://www.kaggle.com/c/aptos2019-blindness-detection/discussion/99603</a></p>",
      "rawMarkdown": "Look for his comments in https://www.kaggle.com/c/aptos2019-blindness-detection/discussion/97860, https://www.kaggle.com/c/aptos2019-blindness-detection/discussion/100815 &amp; in https://www.kaggle.com/c/aptos2019-blindness-detection/discussion/99603",
      "votes": null
    },
    {
      "id": "605324",
      "postDate": "08/22/2019 08:48:32",
      "content": "<p>use the data from previous competition (only train data). Use efficientnet and do augmentations such as zoom,flip,roation. This should help you to achieve 0.8 in LB</p>",
      "rawMarkdown": "use the data from previous competition (only train data). Use efficientnet and do augmentations such as zoom,flip,roation. This should help you to achieve 0.8 in LB",
      "votes": null
    },
    {
      "id": "605327",
      "postDate": "08/22/2019 08:52:04",
      "content": "<p>Thanks. Found him.</p>",
      "rawMarkdown": "Thanks. Found him.",
      "votes": null
    },
    {
      "id": "607746",
      "postDate": "08/25/2019 20:07:35",
      "content": "<p>What is you highest scoring single model? Can you tell that?</p>",
      "rawMarkdown": "What is you highest scoring single model? Can you tell that?",
      "votes": null
    },
    {
      "id": "607757",
      "postDate": "08/25/2019 20:42:32",
      "content": "<p>why do I have this strange feeling it's 0.852 !?</p>",
      "rawMarkdown": "why do I have this strange feeling it's 0.852 !?",
      "votes": null
    },
    {
      "id": "607837",
      "postDate": "08/26/2019 00:57:39",
      "content": "<p>Thanks!!!</p>",
      "rawMarkdown": "Thanks!!!",
      "votes": null
    },
    {
      "id": "611720",
      "postDate": "08/29/2019 11:59:01",
      "content": "<p>Training new data sets with old data sets?I've tried to combine the two, and the loss is around 0.31 and it doesn't come down？</p>",
      "rawMarkdown": "Training new data sets with old data sets?I've tried to combine the two, and the loss is around 0.31 and it doesn't come down？",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 603227,
      "author_name": "dathudeptrai",
      "author_url": "",
      "post_date": "08/20/2019 03:15:10",
      "content": "<p>just do ensemble many and many models =)))</p>",
      "votes": null,
      "replies": [
        {
          "id": 607746,
          "author_name": "nemethpeti",
          "author_url": "",
          "post_date": "08/25/2019 20:07:35",
          "content": "<p>What is you highest scoring single model? Can you tell that?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 607757,
          "author_name": "valanm",
          "author_url": "",
          "post_date": "08/25/2019 20:42:32",
          "content": "<p>why do I have this strange feeling it's 0.852 !?</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 604581,
      "author_name": "abhinand05",
      "author_url": "",
      "post_date": "08/21/2019 14:53:47",
      "content": "<p>Improvements after 0.8? I'm even struggling to hit that mark. Would be nice if you guys can share any tips/tricks/hints. LOL!</p>",
      "votes": null,
      "replies": [
        {
          "id": 604725,
          "author_name": "ravivadapalli",
          "author_url": "",
          "post_date": "08/21/2019 17:37:18",
          "content": "<p>Following the suggestions given by DrHB and Risabh should give you 0.8</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 605119,
          "author_name": "abyaadrafid",
          "author_url": "",
          "post_date": "08/22/2019 04:48:45",
          "content": "<p>I found DrHB's thread but Couldn't find Risabh. Could you link it here?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 605301,
          "author_name": "ravivadapalli",
          "author_url": "",
          "post_date": "08/22/2019 08:35:03",
          "content": "<p>Look for his comments in <a href=\"https://www.kaggle.com/c/aptos2019-blindness-detection/discussion/97860\">https://www.kaggle.com/c/aptos2019-blindness-detection/discussion/97860</a>, <a href=\"https://www.kaggle.com/c/aptos2019-blindness-detection/discussion/100815\">https://www.kaggle.com/c/aptos2019-blindness-detection/discussion/100815</a> &amp; in <a href=\"https://www.kaggle.com/c/aptos2019-blindness-detection/discussion/99603\">https://www.kaggle.com/c/aptos2019-blindness-detection/discussion/99603</a></p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 605327,
          "author_name": "abyaadrafid",
          "author_url": "",
          "post_date": "08/22/2019 08:52:04",
          "content": "<p>Thanks. Found him.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 605023,
      "author_name": "yangfan556677",
      "author_url": "",
      "post_date": "08/22/2019 01:16:18",
      "content": "<p>What method can arrive 0.8, I now 0.78？</p>",
      "votes": null,
      "replies": [
        {
          "id": 605324,
          "author_name": "sabbiracoustic1006",
          "author_url": "",
          "post_date": "08/22/2019 08:48:32",
          "content": "<p>use the data from previous competition (only train data). Use efficientnet and do augmentations such as zoom,flip,roation. This should help you to achieve 0.8 in LB</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 607837,
          "author_name": "yangfan556677",
          "author_url": "",
          "post_date": "08/26/2019 00:57:39",
          "content": "<p>Thanks!!!</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 611720,
          "author_name": "yangfan556677",
          "author_url": "",
          "post_date": "08/29/2019 11:59:01",
          "content": "<p>Training new data sets with old data sets?I've tried to combine the two, and the loss is around 0.31 and it doesn't come down？</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "603193": "Im finding it difficult to improve my LB score after reaching 0.8. Can you guys share any tips/tricks/hints that helped you improve beyond 0.8",
    "603227": "just do ensemble many and many models =)))",
    "604581": "Improvements after 0.8? I'm even struggling to hit that mark. Would be nice if you guys can share any tips/tricks/hints. LOL!",
    "604725": "Following the suggestions given by DrHB and Risabh should give you 0.8",
    "605023": "What method can arrive 0.8, I now 0.78？",
    "605119": "I found DrHB's thread but Couldn't find Risabh. Could you link it here?",
    "605301": "Look for his comments in https://www.kaggle.com/c/aptos2019-blindness-detection/discussion/97860, https://www.kaggle.com/c/aptos2019-blindness-detection/discussion/100815 &amp; in https://www.kaggle.com/c/aptos2019-blindness-detection/discussion/99603",
    "605324": "use the data from previous competition (only train data). Use efficientnet and do augmentations such as zoom,flip,roation. This should help you to achieve 0.8 in LB",
    "605327": "Thanks. Found him.",
    "607746": "What is you highest scoring single model? Can you tell that?",
    "607757": "why do I have this strange feeling it's 0.852 !?",
    "607837": "Thanks!!!",
    "611720": "Training new data sets with old data sets?I've tried to combine the two, and the loss is around 0.31 and it doesn't come down？"
  },
  "source": "meta"
}