{
  "id": 104769,
  "title": "Top Rankers' Solutions (상위 랭커 솔루션 정리)",
  "url": "/competitions/2019-3rd-ml-month-with-kakr/discussion/104769",
  "author_name": "datartist",
  "post_date": "2019-08-19T04:28:26.008000",
  "votes": 8,
  "comment_count": 0,
  "views": 0,
  "content": "<p>안녕하세요:)</p>\n\n<p>먼저 대회 솔루션을 제공해주신 상위 랭커분들께 감사드립니다.\n현재까지 공개된 솔루션들을 간단하게 정리해보았습니다.</p>\n\n<p><a href=\"https://www.kaggle.com/c/2019-3rd-ml-month-with-kakr/discussion/104493#latest-602386\">1st Solution</a> (Public 2nd 0.96604 -&gt; Private 1st 0.96058) '이해찬'\n[discussion] Efficientnet-B7, B6, B5, PNAS, NAS model soft-voting 6-fold tta5 </p>\n\n<p><a href=\"https://www.kaggle.com/yangsaewon/2nd-solution\">2nd Solution</a> (Public 13th 0.96030 -&gt; Private 2nd 0.96033) 'saewonYang'\n[kernel-pytorch] senet142 4fold tta2 + resnext101_32x16d 8-fold tta4</p>\n\n<p><a href=\"https://www.kaggle.com/c/2019-3rd-ml-month-with-kakr/discussion/104509#latest-602383\">7th Solution</a> (Public 1st 0.96652 -&gt; Private 7th 0.95861) 'ildoo Kim'\n[discussion] WSL Network(ResNext 101) + Fast AutoAugment + CutMix</p>\n\n<p><a href=\"https://www.kaggle.com/cruiserx/3rd-ml-month-efficientnet-b3-4-5-ensemble-13th\">13th Solution</a> (Public 6th 0.96317 -&gt; Private 13th 0.95491) 'Steve Jang'\n[kernel-keras] EfficientNet-B3,B4,B5 DNN ensemble  tta10 + MaskRCNN segmentation</p>\n\n<p><a href=\"https://www.kaggle.com/kimtaegwan/14th-solution\">14th Solution</a> (Public 10th 0.96174 -&gt; Private 14th 0.95442) 'Taegwan Kim'\n[discussion] CNN stacking EfficientNetB4, Xception, Resnet50 5-fold </p>\n\n<p><a href=\"https://www.kaggle.com/yseon99/18th-solution-public-0-952-private-0-952\">18th Solution</a>  (Public 27th 0.95361 -&gt; Private 18th 0.95245) 'Yongseon Kim'\n[kernel-keras] ResNext 101 8-fold tta5 + Cutout Augmentation</p>",
  "messages": [
    {
      "id": 602443,
      "postDate": "2019-08-19T04:28:26.010Z",
      "content": "<p>안녕하세요:)</p>\n\n<p>먼저 대회 솔루션을 제공해주신 상위 랭커분들께 감사드립니다.\n현재까지 공개된 솔루션들을 간단하게 정리해보았습니다.</p>\n\n<p><a href=\"https://www.kaggle.com/c/2019-3rd-ml-month-with-kakr/discussion/104493#latest-602386\">1st Solution</a> (Public 2nd 0.96604 -&gt; Private 1st 0.96058) '이해찬'\n[discussion] Efficientnet-B7, B6, B5, PNAS, NAS model soft-voting 6-fold tta5 </p>\n\n<p><a href=\"https://www.kaggle.com/yangsaewon/2nd-solution\">2nd Solution</a> (Public 13th 0.96030 -&gt; Private 2nd 0.96033) 'saewonYang'\n[kernel-pytorch] senet142 4fold tta2 + resnext101_32x16d 8-fold tta4</p>\n\n<p><a href=\"https://www.kaggle.com/c/2019-3rd-ml-month-with-kakr/discussion/104509#latest-602383\">7th Solution</a> (Public 1st 0.96652 -&gt; Private 7th 0.95861) 'ildoo Kim'\n[discussion] WSL Network(ResNext 101) + Fast AutoAugment + CutMix</p>\n\n<p><a href=\"https://www.kaggle.com/cruiserx/3rd-ml-month-efficientnet-b3-4-5-ensemble-13th\">13th Solution</a> (Public 6th 0.96317 -&gt; Private 13th 0.95491) 'Steve Jang'\n[kernel-keras] EfficientNet-B3,B4,B5 DNN ensemble  tta10 + MaskRCNN segmentation</p>\n\n<p><a href=\"https://www.kaggle.com/kimtaegwan/14th-solution\">14th Solution</a> (Public 10th 0.96174 -&gt; Private 14th 0.95442) 'Taegwan Kim'\n[discussion] CNN stacking EfficientNetB4, Xception, Resnet50 5-fold </p>\n\n<p><a href=\"https://www.kaggle.com/yseon99/18th-solution-public-0-952-private-0-952\">18th Solution</a>  (Public 27th 0.95361 -&gt; Private 18th 0.95245) 'Yongseon Kim'\n[kernel-keras] ResNext 101 8-fold tta5 + Cutout Augmentation</p>",
      "rawMarkdown": "안녕하세요:)\n\n먼저 대회 솔루션을 제공해주신 상위 랭커분들께 감사드립니다.\n현재까지 공개된 솔루션들을 간단하게 정리해보았습니다.\n\n[1st Solution](https://www.kaggle.com/c/2019-3rd-ml-month-with-kakr/discussion/104493#latest-602386) (Public 2nd 0.96604 -&gt; Private 1st 0.96058) '이해찬'\n[discussion] Efficientnet-B7, B6, B5, PNAS, NAS model soft-voting 6-fold tta5 \n\n[2nd Solution](https://www.kaggle.com/yangsaewon/2nd-solution) (Public 13th 0.96030 -&gt; Private 2nd 0.96033) 'saewonYang'\n[kernel-pytorch] senet142 4fold tta2 + resnext101_32x16d 8-fold tta4\n\n[7th Solution](https://www.kaggle.com/c/2019-3rd-ml-month-with-kakr/discussion/104509#latest-602383) (Public 1st 0.96652 -&gt; Private 7th 0.95861) 'ildoo Kim'\n[discussion] WSL Network(ResNext 101) + Fast AutoAugment + CutMix\n\n[13th Solution](https://www.kaggle.com/cruiserx/3rd-ml-month-efficientnet-b3-4-5-ensemble-13th) (Public 6th 0.96317 -&gt; Private 13th 0.95491) 'Steve Jang'\n[kernel-keras] EfficientNet-B3,B4,B5 DNN ensemble  tta10 + MaskRCNN segmentation\n\n[14th Solution](https://www.kaggle.com/kimtaegwan/14th-solution) (Public 10th 0.96174 -&gt; Private 14th 0.95442) 'Taegwan Kim'\n[discussion] CNN stacking EfficientNetB4, Xception, Resnet50 5-fold \n\n[18th Solution](https://www.kaggle.com/yseon99/18th-solution-public-0-952-private-0-952)  (Public 27th 0.95361 -&gt; Private 18th 0.95245) 'Yongseon Kim'\n[kernel-keras] ResNext 101 8-fold tta5 + Cutout Augmentation\n",
      "votes": 8
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "602443": "안녕하세요:)\n\n먼저 대회 솔루션을 제공해주신 상위 랭커분들께 감사드립니다.\n현재까지 공개된 솔루션들을 간단하게 정리해보았습니다.\n\n[1st Solution](https://www.kaggle.com/c/2019-3rd-ml-month-with-kakr/discussion/104493#latest-602386) (Public 2nd 0.96604 -&gt; Private 1st 0.96058) '이해찬'\n[discussion] Efficientnet-B7, B6, B5, PNAS, NAS model soft-voting 6-fold tta5 \n\n[2nd Solution](https://www.kaggle.com/yangsaewon/2nd-solution) (Public 13th 0.96030 -&gt; Private 2nd 0.96033) 'saewonYang'\n[kernel-pytorch] senet142 4fold tta2 + resnext101_32x16d 8-fold tta4\n\n[7th Solution](https://www.kaggle.com/c/2019-3rd-ml-month-with-kakr/discussion/104509#latest-602383) (Public 1st 0.96652 -&gt; Private 7th 0.95861) 'ildoo Kim'\n[discussion] WSL Network(ResNext 101) + Fast AutoAugment + CutMix\n\n[13th Solution](https://www.kaggle.com/cruiserx/3rd-ml-month-efficientnet-b3-4-5-ensemble-13th) (Public 6th 0.96317 -&gt; Private 13th 0.95491) 'Steve Jang'\n[kernel-keras] EfficientNet-B3,B4,B5 DNN ensemble  tta10 + MaskRCNN segmentation\n\n[14th Solution](https://www.kaggle.com/kimtaegwan/14th-solution) (Public 10th 0.96174 -&gt; Private 14th 0.95442) 'Taegwan Kim'\n[discussion] CNN stacking EfficientNetB4, Xception, Resnet50 5-fold \n\n[18th Solution](https://www.kaggle.com/yseon99/18th-solution-public-0-952-private-0-952)  (Public 27th 0.95361 -&gt; Private 18th 0.95245) 'Yongseon Kim'\n[kernel-keras] ResNext 101 8-fold tta5 + Cutout Augmentation\n"
  }
}