{
  "id": 27880,
  "title": "Post of thanks",
  "url": "/competitions/outbrain-click-prediction/discussion/27880",
  "author_name": "",
  "post_date": "2017-01-18T21:45:20.283Z",
  "votes": 12,
  "comment_count": 12,
  "views": 568,
  "content": "<p>Hey there,</p>\n\n<p>Thanks to Kaggle and Outbrain for the competition, and thanks to all who participated, it was a pleasure to assess skills here.</p>\n\n<p>Looking forward for shared solutions from the winners, congrats to them in advance :-)</p>\n\n<p>Andrii</p>",
  "messages": [
    {
      "id": "157002",
      "postDate": "01/18/2017 21:45:20",
      "content": "<p>Hey there,</p>\n\n<p>Thanks to Kaggle and Outbrain for the competition, and thanks to all who participated, it was a pleasure to assess skills here.</p>\n\n<p>Looking forward for shared solutions from the winners, congrats to them in advance :-)</p>\n\n<p>Andrii</p>",
      "rawMarkdown": "Hey there,\r\n\r\nThanks to Kaggle and Outbrain for the competition, and thanks to all who participated, it was a pleasure to assess skills here.\r\n\r\nLooking forward for shared solutions from the winners, congrats to them in advance :-)\r\n\r\nAndrii",
      "votes": null
    },
    {
      "id": "157004",
      "postDate": "01/18/2017 21:49:19",
      "content": "<p>Thank you Andrii! It is really impressive for your solo play. Seems we all finished our business. Good luck to top 2!</p>",
      "rawMarkdown": "Thank you Andrii! It is really impressive for your solo play. Seems we all finished our business. Good luck to top 2!",
      "votes": null
    },
    {
      "id": "157010",
      "postDate": "01/18/2017 22:15:20",
      "content": "<p>It has been a good challenge. :-)</p>\n\n<p>Will be interesting to hear more about the solutions. I assume a lot of libffm's have been used. I got it working pretty well on a subset of the data, but not on the full training set. </p>",
      "rawMarkdown": "It has been a good challenge. :-)\r\n\r\nWill be interesting to hear more about the solutions. I assume a lot of libffm's have been used. I got it working pretty well on a subset of the data, but not on the full training set.",
      "votes": null
    },
    {
      "id": "157018",
      "postDate": "01/18/2017 22:55:09",
      "content": "<p>Top 2 are so crazy. Improve everyday...</p>",
      "rawMarkdown": "Top 2 are so crazy. Improve everyday...",
      "votes": null
    },
    {
      "id": "157025",
      "postDate": "01/18/2017 23:17:54",
      "content": "<p>This was my first competition! I learned so much over this months. I'm looking forward for the winners solutions! </p>",
      "rawMarkdown": "This was my first competition! I learned so much over this months. I'm looking forward for the winners solutions!",
      "votes": null
    },
    {
      "id": "157031",
      "postDate": "01/19/2017 00:06:22",
      "content": "<p>This was such an awesome competition. Congrats to the winners! This really forced me to think out of the box in terms of data management and was also the first time I have used FTRL, Vowpal Wabbit and FFM. </p>",
      "rawMarkdown": "This was such an awesome competition. Congrats to the winners! This really forced me to think out of the box in terms of data management and was also the first time I have used FTRL, Vowpal Wabbit and FFM.",
      "votes": null
    },
    {
      "id": "157033",
      "postDate": "01/19/2017 00:08:28",
      "content": "<p>Yeah, thanks to all participants and organizers! It was really fun!</p>",
      "rawMarkdown": "Yeah, thanks to all participants and organizers! It was really fun!",
      "votes": null
    },
    {
      "id": "157035",
      "postDate": "01/19/2017 00:12:00",
      "content": "<p>Thanks for Kaggle and Outbrain for this great challenge!\nCongratulations to the winners: code monkey, brain-afk and Three Data Points.\nIt was a hard competition. We just realized how to beat 0.69+ today ;-P\nEager to read winners solutions.\nTchau</p>",
      "rawMarkdown": "Thanks for Kaggle and Outbrain for this great challenge!\r\nCongratulations to the winners: code monkey, brain-afk and Three Data Points.\r\nIt was a hard competition. We just realized how to beat 0.69+ today ;-P\r\nEager to read winners solutions.\r\nTchau",
      "votes": null
    },
    {
      "id": "157103",
      "postDate": "01/19/2017 08:01:11",
      "content": "<p>@Gilberto. And how to beat it? Can you share your idea briefly? Does it use custome features?</p>",
      "rawMarkdown": "Gilberto. And how to beat it? Can you share your idea briefly? Does it use custome features?",
      "votes": null
    },
    {
      "id": "157137",
      "postDate": "01/19/2017 11:54:59",
      "content": "<p>@Evgeny_Semyonov there are some features from Views that helps a lot. First one is the LEAK. Second are some lead/lag features from Views based in UUID.  Actually lead features (or next) performed better than lag. ie:  Next Document_id per UUID, next topic per UUID, etc... That features can help break 0.69 map. Actually my team mate Ash found that on 17-01. The model that finished training 5min after competition deadline would placed us #8  ;-P</p>",
      "rawMarkdown": "Evgeny_Semyonov there are some features from Views that helps a lot. First one is the LEAK. Second are some lead/lag features from Views based in UUID.  Actually lead features (or next) performed better than lag. ie:  Next Document_id per UUID, next topic per UUID, etc... That features can help break 0.69 map. Actually my team mate Ash found that on 17-01. The model that finished training 5min after competition deadline would placed us #8  ;-P",
      "votes": null
    },
    {
      "id": "157177",
      "postDate": "01/19/2017 15:02:23",
      "content": "<p>Congrats to the winners! It's great to see new frameworks being developed just for this competition @rcarson. This competition was very special for me, as the first I was really engaged, and could end up in the first LB page scroll (19th) \\o/ . @RDizzl3, like you, my first time using FTRL, VW, FFM and LightGBM. Lot's of learning on your posts on forums, kernels and past solutions. Kaggle is a great community. Thanks!</p>",
      "rawMarkdown": "Congrats to the winners! It's great to see new frameworks being developed just for this competition @rcarson. This competition was very special for me, as the first I was really engaged, and could end up in the first LB page scroll (19th) \\o/ . @RDizzl3, like you, my first time using FTRL, VW, FFM and LightGBM. Lot's of learning on your posts on forums, kernels and past solutions. Kaggle is a great community. Thanks!",
      "votes": null
    },
    {
      "id": "157202",
      "postDate": "01/19/2017 16:50:06",
      "content": "<p>Congratulations to everyone, that was very nice to follow the fight between top teams at the top of LB!\nTime to read all the great solutions threads :)</p>",
      "rawMarkdown": "Congratulations to everyone, that was very nice to follow the fight between top teams at the top of LB!\r\nTime to read all the great solutions threads :)",
      "votes": null
    },
    {
      "id": "157410",
      "postDate": "01/20/2017 21:23:33",
      "content": "<p>Congratulations to the winners, and thanks to Kaggle and Outbrain for a really interesting and challenging competition.</p>",
      "rawMarkdown": "Congratulations to the winners, and thanks to Kaggle and Outbrain for a really interesting and challenging competition.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 157004,
      "author_name": "jiweiliu",
      "author_url": "",
      "post_date": "01/18/2017 21:49:19",
      "content": "<p>Thank you Andrii! It is really impressive for your solo play. Seems we all finished our business. Good luck to top 2!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 157010,
      "author_name": "frederik",
      "author_url": "",
      "post_date": "01/18/2017 22:15:20",
      "content": "<p>It has been a good challenge. :-)</p>\n\n<p>Will be interesting to hear more about the solutions. I assume a lot of libffm's have been used. I got it working pretty well on a subset of the data, but not on the full training set. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 157018,
      "author_name": "beedata",
      "author_url": "",
      "post_date": "01/18/2017 22:55:09",
      "content": "<p>Top 2 are so crazy. Improve everyday...</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 157025,
      "author_name": "guilhermesantos",
      "author_url": "",
      "post_date": "01/18/2017 23:17:54",
      "content": "<p>This was my first competition! I learned so much over this months. I'm looking forward for the winners solutions! </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 157031,
      "author_name": "rdizzl3",
      "author_url": "",
      "post_date": "01/19/2017 00:06:22",
      "content": "<p>This was such an awesome competition. Congrats to the winners! This really forced me to think out of the box in terms of data management and was also the first time I have used FTRL, Vowpal Wabbit and FFM. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 157033,
      "author_name": "alexeynoskov",
      "author_url": "",
      "post_date": "01/19/2017 00:08:28",
      "content": "<p>Yeah, thanks to all participants and organizers! It was really fun!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 157035,
      "author_name": "titericz",
      "author_url": "",
      "post_date": "01/19/2017 00:12:00",
      "content": "<p>Thanks for Kaggle and Outbrain for this great challenge!\nCongratulations to the winners: code monkey, brain-afk and Three Data Points.\nIt was a hard competition. We just realized how to beat 0.69+ today ;-P\nEager to read winners solutions.\nTchau</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 157103,
      "author_name": "lightforever",
      "author_url": "",
      "post_date": "01/19/2017 08:01:11",
      "content": "<p>@Gilberto. And how to beat it? Can you share your idea briefly? Does it use custome features?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 157137,
      "author_name": "titericz",
      "author_url": "",
      "post_date": "01/19/2017 11:54:59",
      "content": "<p>@Evgeny_Semyonov there are some features from Views that helps a lot. First one is the LEAK. Second are some lead/lag features from Views based in UUID.  Actually lead features (or next) performed better than lag. ie:  Next Document_id per UUID, next topic per UUID, etc... That features can help break 0.69 map. Actually my team mate Ash found that on 17-01. The model that finished training 5min after competition deadline would placed us #8  ;-P</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 157177,
      "author_name": "gspmoreira",
      "author_url": "",
      "post_date": "01/19/2017 15:02:23",
      "content": "<p>Congrats to the winners! It's great to see new frameworks being developed just for this competition @rcarson. This competition was very special for me, as the first I was really engaged, and could end up in the first LB page scroll (19th) \\o/ . @RDizzl3, like you, my first time using FTRL, VW, FFM and LightGBM. Lot's of learning on your posts on forums, kernels and past solutions. Kaggle is a great community. Thanks!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 157202,
      "author_name": "phansoks",
      "author_url": "",
      "post_date": "01/19/2017 16:50:06",
      "content": "<p>Congratulations to everyone, that was very nice to follow the fight between top teams at the top of LB!\nTime to read all the great solutions threads :)</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 157410,
      "author_name": "dslate",
      "author_url": "",
      "post_date": "01/20/2017 21:23:33",
      "content": "<p>Congratulations to the winners, and thanks to Kaggle and Outbrain for a really interesting and challenging competition.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "157002": "Hey there,\r\n\r\nThanks to Kaggle and Outbrain for the competition, and thanks to all who participated, it was a pleasure to assess skills here.\r\n\r\nLooking forward for shared solutions from the winners, congrats to them in advance :-)\r\n\r\nAndrii",
    "157004": "Thank you Andrii! It is really impressive for your solo play. Seems we all finished our business. Good luck to top 2!",
    "157010": "It has been a good challenge. :-)\r\n\r\nWill be interesting to hear more about the solutions. I assume a lot of libffm's have been used. I got it working pretty well on a subset of the data, but not on the full training set.",
    "157018": "Top 2 are so crazy. Improve everyday...",
    "157025": "This was my first competition! I learned so much over this months. I'm looking forward for the winners solutions!",
    "157031": "This was such an awesome competition. Congrats to the winners! This really forced me to think out of the box in terms of data management and was also the first time I have used FTRL, Vowpal Wabbit and FFM.",
    "157033": "Yeah, thanks to all participants and organizers! It was really fun!",
    "157035": "Thanks for Kaggle and Outbrain for this great challenge!\r\nCongratulations to the winners: code monkey, brain-afk and Three Data Points.\r\nIt was a hard competition. We just realized how to beat 0.69+ today ;-P\r\nEager to read winners solutions.\r\nTchau",
    "157103": "Gilberto. And how to beat it? Can you share your idea briefly? Does it use custome features?",
    "157137": "Evgeny_Semyonov there are some features from Views that helps a lot. First one is the LEAK. Second are some lead/lag features from Views based in UUID.  Actually lead features (or next) performed better than lag. ie:  Next Document_id per UUID, next topic per UUID, etc... That features can help break 0.69 map. Actually my team mate Ash found that on 17-01. The model that finished training 5min after competition deadline would placed us #8  ;-P",
    "157177": "Congrats to the winners! It's great to see new frameworks being developed just for this competition @rcarson. This competition was very special for me, as the first I was really engaged, and could end up in the first LB page scroll (19th) \\o/ . @RDizzl3, like you, my first time using FTRL, VW, FFM and LightGBM. Lot's of learning on your posts on forums, kernels and past solutions. Kaggle is a great community. Thanks!",
    "157202": "Congratulations to everyone, that was very nice to follow the fight between top teams at the top of LB!\r\nTime to read all the great solutions threads :)",
    "157410": "Congratulations to the winners, and thanks to Kaggle and Outbrain for a really interesting and challenging competition."
  },
  "source": "meta"
}