{
  "id": 15463,
  "title": "Score 0.981817 with a simple gbm",
  "url": "/competitions/flavours-of-physics/discussion/15463",
  "author_name": "",
  "post_date": "2015-07-23T02:13:55.587Z",
  "votes": 6,
  "comment_count": 8,
  "views": 1506,
  "content": "<p><a href=\"http://pastebin.com/sJLUkMwq\">http://pastebin.com/sJLUkMwq</a></p>\n\n<p>because if you can't beat them, join them?</p>",
  "messages": [
    {
      "id": "86559",
      "postDate": "07/23/2015 02:13:55",
      "content": "<p><a href=\"http://pastebin.com/sJLUkMwq\">http://pastebin.com/sJLUkMwq</a></p>\n\n<p>because if you can't beat them, join them?</p>",
      "rawMarkdown": "http://pastebin.com/sJLUkMwq\r\n\r\nbecause if you can't beat them, join them?",
      "votes": null
    },
    {
      "id": "86567",
      "postDate": "07/23/2015 04:46:34",
      "content": "<p>Great work, this is going to be a fun 0.018183 race</p>",
      "rawMarkdown": "Great work, this is going to be a fun 0.018183 race",
      "votes": null
    },
    {
      "id": "86595",
      "postDate": "07/23/2015 10:13:04",
      "content": "<p>Come on James, copy it to the Script section. You know you want to really! (ps I've &quot;clicked and submitted &quot; twice today already :O)</p>",
      "rawMarkdown": "Come on James, copy it to the Script section. You know you want to really! (ps I've \"clicked and submitted \" twice today already :O)",
      "votes": null
    },
    {
      "id": "86601",
      "postDate": "07/23/2015 11:21:37",
      "content": "<p>[quote=Bluefool;86595]</p>\n\n<p>Come on James, copy it to the Script section. You know you want to really! (ps I've &quot;clicked and submitted &quot; twice today already :O)</p>\n\n<p>[/quote]</p>\n\n<p>It's already been beaten by an average of gbm and random forest posted by Ben Hamner. Maybe I'll try adding a neural network component.</p>",
      "rawMarkdown": "[quote=Bluefool;86595]\r\n\r\nCome on James, copy it to the Script section. You know you want to really! (ps I've \"clicked and submitted \" twice today already :O)\r\n\r\n[/quote]\r\n\r\nIt's already been beaten by an average of gbm and random forest posted by Ben Hamner. Maybe I'll try adding a neural network component.",
      "votes": null
    },
    {
      "id": "86605",
      "postDate": "07/23/2015 11:45:34",
      "content": "<p>it improved my score!</p>",
      "rawMarkdown": "it improved my score!",
      "votes": null
    },
    {
      "id": "86624",
      "postDate": "07/23/2015 13:55:18",
      "content": "<p>Code there does not give me D statistics less than 0.09, am I missing something here? I am using ks.test from dgof package, as follows\nKSmetric &lt;- ks.test(agreement[which(agreement$signal == 0),]$prob,\n                    unlist(agreement[which(agreement$signal == 1),]$prob),\n                    simulate.p.value=FALSE)</p>",
      "rawMarkdown": "Code there does not give me D statistics less than 0.09, am I missing something here? I am using ks.test from dgof package, as follows\r\nKSmetric <- ks.test(agreement[which(agreement$signal == 0),]$prob,\r\n                    unlist(agreement[which(agreement$signal == 1),]$prob),\r\n                    simulate.p.value=FALSE)",
      "votes": null
    },
    {
      "id": "86629",
      "postDate": "07/23/2015 14:19:53",
      "content": "<p>[quote=xinus;86624]</p>\n\n<p>Code there does not give me D statistics less than 0.09...</p>\n\n<p>[/quote]</p>\n\n<p>I did not run the KS test, I just submitted and confirmed that I did not get an error.</p>",
      "rawMarkdown": "[quote=xinus;86624]\r\n\r\nCode there does not give me D statistics less than 0.09...\r\n\r\n[/quote]\r\n\r\nI did not run the KS test, I just submitted and confirmed that I did not get an error.",
      "votes": null
    },
    {
      "id": "86644",
      "postDate": "07/23/2015 15:03:32",
      "content": "<p>[quote=xinus;86624]</p>\n\n<p>Code there does not give me D statistics less than 0.09, am I missing something here? I am using ks.test from dgof package, as follows\nKSmetric &lt;- ks.test(agreement[which(agreement$signal == 0),]$prob,\n                    unlist(agreement[which(agreement$signal == 1),]$prob),\n                    simulate.p.value=FALSE)</p>\n\n<p>[/quote]</p>\n\n<p>The random seed is not set so you are getting different random shuffles from the original script.</p>",
      "rawMarkdown": "[quote=xinus;86624]\r\n\r\nCode there does not give me D statistics less than 0.09, am I missing something here? I am using ks.test from dgof package, as follows\r\nKSmetric <- ks.test(agreement[which(agreement$signal == 0),]$prob,\r\n                    unlist(agreement[which(agreement$signal == 1),]$prob),\r\n                    simulate.p.value=FALSE)\r\n\r\n\r\n[/quote]\r\n\r\nThe random seed is not set so you are getting different random shuffles from the original script.",
      "votes": null
    },
    {
      "id": "86672",
      "postDate": "07/23/2015 18:22:57",
      "content": "<p>Huh, sorry I use python not R, but don't you need to normalise/whiten your data before you do that? Is it automatically whitened inside of gbm?</p>",
      "rawMarkdown": "Huh, sorry I use python not R, but don't you need to normalise/whiten your data before you do that? Is it automatically whitened inside of gbm?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 86567,
      "author_name": "carloshuertas",
      "author_url": "",
      "post_date": "07/23/2015 04:46:34",
      "content": "<p>Great work, this is going to be a fun 0.018183 race</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 86595,
      "author_name": "domcastro",
      "author_url": "",
      "post_date": "07/23/2015 10:13:04",
      "content": "<p>Come on James, copy it to the Script section. You know you want to really! (ps I've &quot;clicked and submitted &quot; twice today already :O)</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 86601,
      "author_name": "jfkingiii",
      "author_url": "",
      "post_date": "07/23/2015 11:21:37",
      "content": "<p>[quote=Bluefool;86595]</p>\n\n<p>Come on James, copy it to the Script section. You know you want to really! (ps I've &quot;clicked and submitted &quot; twice today already :O)</p>\n\n<p>[/quote]</p>\n\n<p>It's already been beaten by an average of gbm and random forest posted by Ben Hamner. Maybe I'll try adding a neural network component.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 86605,
      "author_name": "domcastro",
      "author_url": "",
      "post_date": "07/23/2015 11:45:34",
      "content": "<p>it improved my score!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 86624,
      "author_name": "xinusx",
      "author_url": "",
      "post_date": "07/23/2015 13:55:18",
      "content": "<p>Code there does not give me D statistics less than 0.09, am I missing something here? I am using ks.test from dgof package, as follows\nKSmetric &lt;- ks.test(agreement[which(agreement$signal == 0),]$prob,\n                    unlist(agreement[which(agreement$signal == 1),]$prob),\n                    simulate.p.value=FALSE)</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 86629,
      "author_name": "jfkingiii",
      "author_url": "",
      "post_date": "07/23/2015 14:19:53",
      "content": "<p>[quote=xinus;86624]</p>\n\n<p>Code there does not give me D statistics less than 0.09...</p>\n\n<p>[/quote]</p>\n\n<p>I did not run the KS test, I just submitted and confirmed that I did not get an error.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 86644,
      "author_name": "mtutek",
      "author_url": "",
      "post_date": "07/23/2015 15:03:32",
      "content": "<p>[quote=xinus;86624]</p>\n\n<p>Code there does not give me D statistics less than 0.09, am I missing something here? I am using ks.test from dgof package, as follows\nKSmetric &lt;- ks.test(agreement[which(agreement$signal == 0),]$prob,\n                    unlist(agreement[which(agreement$signal == 1),]$prob),\n                    simulate.p.value=FALSE)</p>\n\n<p>[/quote]</p>\n\n<p>The random seed is not set so you are getting different random shuffles from the original script.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 86672,
      "author_name": "erlend",
      "author_url": "",
      "post_date": "07/23/2015 18:22:57",
      "content": "<p>Huh, sorry I use python not R, but don't you need to normalise/whiten your data before you do that? Is it automatically whitened inside of gbm?</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "86559": "http://pastebin.com/sJLUkMwq\r\n\r\nbecause if you can't beat them, join them?",
    "86567": "Great work, this is going to be a fun 0.018183 race",
    "86595": "Come on James, copy it to the Script section. You know you want to really! (ps I've \"clicked and submitted \" twice today already :O)",
    "86601": "[quote=Bluefool;86595]\r\n\r\nCome on James, copy it to the Script section. You know you want to really! (ps I've \"clicked and submitted \" twice today already :O)\r\n\r\n[/quote]\r\n\r\nIt's already been beaten by an average of gbm and random forest posted by Ben Hamner. Maybe I'll try adding a neural network component.",
    "86605": "it improved my score!",
    "86624": "Code there does not give me D statistics less than 0.09, am I missing something here? I am using ks.test from dgof package, as follows\r\nKSmetric <- ks.test(agreement[which(agreement$signal == 0),]$prob,\r\n                    unlist(agreement[which(agreement$signal == 1),]$prob),\r\n                    simulate.p.value=FALSE)",
    "86629": "[quote=xinus;86624]\r\n\r\nCode there does not give me D statistics less than 0.09...\r\n\r\n[/quote]\r\n\r\nI did not run the KS test, I just submitted and confirmed that I did not get an error.",
    "86644": "[quote=xinus;86624]\r\n\r\nCode there does not give me D statistics less than 0.09, am I missing something here? I am using ks.test from dgof package, as follows\r\nKSmetric <- ks.test(agreement[which(agreement$signal == 0),]$prob,\r\n                    unlist(agreement[which(agreement$signal == 1),]$prob),\r\n                    simulate.p.value=FALSE)\r\n\r\n\r\n[/quote]\r\n\r\nThe random seed is not set so you are getting different random shuffles from the original script.",
    "86672": "Huh, sorry I use python not R, but don't you need to normalise/whiten your data before you do that? Is it automatically whitened inside of gbm?"
  },
  "source": "meta"
}