{
  "id": 5289,
  "title": "Start uploading your code",
  "url": "/competitions/cause-effect-pairs/discussion/5289",
  "author_name": "",
  "post_date": "2013-08-04T20:56:54.890Z",
  "votes": null,
  "comment_count": 29,
  "views": 7780,
  "content": "<p>Only 2 more weeks until the deadline. The top ranking participants should start uploading their code. Avoid last minute rushes and allow us to take a peek at your code to make sure we can run it. You can keep updating it until the deadline.</p>\n<p>Instructions to prepare your code are found at:</p>\n<p>http://www.causality.inf.ethz.ch/cause-effect.php?page=tasks#software (R code is also OK).</p>\n<p>Notify us by email at causality at chalearn dot org when code is uploaded.</p>\n<p>Thanks!</p>",
  "messages": [
    {
      "id": "28184",
      "postDate": "08/04/2013 20:56:54",
      "content": "<p>Only 2 more weeks until the deadline. The top ranking participants should start uploading their code. Avoid last minute rushes and allow us to take a peek at your code to make sure we can run it. You can keep updating it until the deadline.</p>\n<p>Instructions to prepare your code are found at:</p>\n<p>http://www.causality.inf.ethz.ch/cause-effect.php?page=tasks#software (R code is also OK).</p>\n<p>Notify us by email at causality at chalearn dot org when code is uploaded.</p>\n<p>Thanks!</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "28654",
      "postDate": "08/14/2013 00:37:24",
      "content": "<p>Thank you very much to those who started uploading their code. The others should REALLY do this as soon as possible. Remember:</p>\n<p>- Name the archive after your team name</p>\n<p>- Include ONE up-to-date README file with</p>\n<p style=\"padding-left: 30px\">* You team name</p>\n<p style=\"padding-left: 30px\">* A contact EMAIL</p>\n<p style=\"padding-left: 30px\">* SIMPLE and complete instructions (including ALL the software dependencies)</p>\n<p>- Do NOT include a copy of the whole dataset as part of your submission. Add a SMALL data sample and a SMALL test (running is a few seconds) so the code can be checked rapidly</p>\n<p>- If training software is included (optional) include instructions to process the final test data WITHOUT retraining (i.e. include a TRAINED MODEL for which predict can just be run without retraining.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "28656",
      "postDate": "08/14/2013 02:06:48",
      "content": "<p>can I assume that all the csv data files are available by default ? and is the .mat file valid for the submission purpose ?&nbsp;</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "28657",
      "postDate": "08/14/2013 02:17:27",
      "content": "<p>The final test data files will be called&nbsp;CEfinal_test_pairs.csv and&nbsp;CEfinal_test_publicinfo.csv. For the split format, there will be a file&nbsp;CEfinal_test_split, with files test1.txt, test2.txr, etc.</p>\n<p>You don't need to attach data to your code (unless you want to provide a small sample to test purposes).</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "28677",
      "postDate": "08/14/2013 11:32:40",
      "content": "<p>Hello,</p>\n<p><span style=\"line-height: 1.4\">Can we assume that your Python environment already contains the following typical libraries?</span></p>\n<ul>\n<li>numpy</li>\n<li>pandas</li>\n<li>scipy</li>\n<li>sklearn</li>\n<li>pylab</li>\n</ul>\n<p>Thank you.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "28700",
      "postDate": "08/14/2013 17:06:26",
      "content": "<p>You can indicate which library you use and the version and we can try to install the libraries and check whether everything works fine. But this is at your own risks. I recommend that you provide a piece of test code we can run that checks the full installation by training and testing on a small data subset and checks the values obtained are identical to those produced when you run the code on your computer. In particular, make sure that you do not depend on randomly generated numbers that change your result. Draw random numbers in advance and always use the same values.</p>\n<p>Note: for verification, we do not require retraining, so consider re-loading a trained model and using it to make predictions on the test data.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "28728",
      "postDate": "08/15/2013 12:47:41",
      "content": "<p>Is it ok for the program to print some statistics or timing information that doesn't stay the same every run? I.e. how many seconds/milliseconds it took to read a file, do analysis 1, do analysis 2 etc..</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "28732",
      "postDate": "08/15/2013 13:34:27",
      "content": "<p>Yes. It is actually a good idea to print some diagnostics.&nbsp;</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "28734",
      "postDate": "08/15/2013 13:38:05",
      "content": "<p>I think I won't be able to improve much above 0.75, I tried to dig more features out of the scatter data but currently I am using only one RF regressor. Could any one hint some light on improving further the results using some more **advanced** regressors ?&nbsp;</p>\n<p>And I am very keen to see the method that was used to reach Rank1!</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "28736",
      "postDate": "08/15/2013 14:26:37",
      "content": "<p>[quote=liubenyuan;28734]</p>\n<p>I think I won't be able to improve much above 0.75, I tried to dig more features out of the scatter data but currently I am using only one RF regressor. Could any one hint some light on improving further the results using some more **advanced** regressors ?&nbsp;</p>\n<p>And I am very keen to see the method that was used to reach Rank1!</p>\n<p>[/quote]</p>\n<p>&nbsp;</p>\n<p>did you also use the supplimentary data?</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "28738",
      "postDate": "08/15/2013 14:54:16",
      "content": "<p>Yes, both SUP1 and SUP2 are used.&nbsp;</p>\n<p>I found that the CV test on SUP2 provided a good match to the final valid dataset.&nbsp;</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "28739",
      "postDate": "08/15/2013 14:56:54",
      "content": "<p>[quote=liubenyuan;28738]</p>\n<p>Yes, both SUP1 and SUP2 are used.&nbsp;</p>\n<p>I found that the CV test on SUP2 provided a good match to the final valid dataset.&nbsp;</p>\n<p>[/quote]</p>\n<p>&nbsp;</p>\n<p>what about SUP3? any specific reason for not using it? would you mind telling on how many features you are training your model?</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "28740",
      "postDate": "08/15/2013 15:02:35",
      "content": "<p>81 features, most of them are from the python code. and to those who use matlab :</p>\n<p>1. ANM, ANMD, PNL are far too slow to be used to extract the features. and ANM and ANMD reported NaN results on some pairs (warning).</p>\n<p>2. LINGAM is fast, however the result on the quantized data are not so good.</p>\n<p>3. the nips-code folder in GPI is very valuable, and I think the features extracted using this method might be good for training. However they are relatively too slow running in my laptop ( ~approximately 7 days on the whole train, SUP1, SUP2 datasets). some one may give it a try !!</p>\n<p>4. IGCI is extremely fast, some of my other features are generated using variant versions of this code.&nbsp;</p>\n<p>5. and some fabricated features by intuition of viewing (ax,bx) paris.</p>\n<p>&nbsp;</p>\n<p>It is a great challenge. I hope these information might help in these days.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "28751",
      "postDate": "08/15/2013 18:26:25",
      "content": "<p>I also had good success with the Lingam algorithms. I have only 27 features left after I did a round of feature selection recently. I tried additive noise models, but could not get good results from them. I tried training on each target value one at a time and then using the results of that as features for a second round of testing, but again the results were worse if anything. &nbsp;I found IGCI worked well on the initial data but not so well on the 'fixed' data.</p>\n<p>&nbsp;</p>\n<p>By some last minute blending of models I've improved my score by 0.01, but I suspect to go further I'd need to create some new features (ie get PNL, ANM or something new working).</p>\n<p>&nbsp;</p>\n<p>My 10-fold CV score is actually 0.76 (compared to 0.72 on the Leaderboard). How do other people's CV scores compare to the leaderboard?&nbsp;</p>\n<p>I've learnt an incredible amount from this competition - thank you Isabelle and Kaggle.</p>\n<p>&nbsp;</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "28766",
      "postDate": "08/15/2013 21:04:48",
      "content": "<p>Hi Eoin,</p>\n<p>&nbsp; &nbsp;My CV and leaderboard scores have actually been quite consistent. For my latest submissions, I got&nbsp;0.701680 CV vs&nbsp;0.71140 leaderboard. Others have been similarly close, usually within the RMS of the CV scores.</p>\n<p>&nbsp; &nbsp;At one point, my classifier was somewhat asymmetric, performing better for backward AUC than for forward AUC. CV scores on those models tended to have a larger deviance with respect to the leaderboard scores, possibly because of differences in sample composition. Are your models all close to symmetric?</p>\n<p>&nbsp;</p>\n<p>- Emanuel</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "28768",
      "postDate": "08/15/2013 21:15:07",
      "content": "<p>Hi Emmanuel,</p>\n<p>I think you're right there - my models are giving asymmetric scores, with a difference of almost 0.1. Interesting... probably too late to change my code, but not too late to retrain ;)</p>\n<p>&nbsp;</p>\n<p>Eoin</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "28781",
      "postDate": "08/16/2013 01:24:39",
      "content": "<p>Well, I wish you good luck (but not too much luck ;) )&nbsp;</p>\n<p>&nbsp;</p>\n<p>- Emanuel</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "28798",
      "postDate": "08/16/2013 07:47:33",
      "content": "<p>I fixed an asymmetry in my model and my score improved slightly. Still searching for the magic sauce that the people above 0.8 are using...</p>\n<p>&nbsp;</p>\n<p>Eoin</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "28799",
      "postDate": "08/16/2013 07:50:52",
      "content": "<p>[quote=Eoin Lawless;28798]</p>\n<p>I fixed an asymmetry in my model and my score improved slightly. Still searching for the magic sauce that the people above 0.8 are using...</p>\n<p>&nbsp;</p>\n<p>Eoin</p>\n<p>[/quote]</p>\n<p>&nbsp;</p>\n<p>damn, u took our position..</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "28800",
      "postDate": "08/16/2013 08:53:54",
      "content": "<p>I'm sure there'll be a lot of activity over the next few days... I don't expect to hold it for long...</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "28801",
      "postDate": "08/16/2013 08:55:36",
      "content": "<p>[quote=Eoin Lawless;28800]</p>\n<p>I'm sure there'll be a lot of activity over the next few days... I don't expect to hold it for long...</p>\n<p>[/quote]</p>\n<p>&nbsp;</p>\n<p>yeah, seems so. new teams coming up and huge increase in ranks and all...</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "28804",
      "postDate": "08/16/2013 09:21:32",
      "content": "<p>yeah, it was also exciting to know handling the asymmetric could improve another 0.02 :P</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "28806",
      "postDate": "08/16/2013 09:34:42",
      "content": "<p>I only got 0.011 from the fix, but my CV scores are still higher than the leaderboard score so I may still be missing something.&nbsp;</p>\n<p><span style=\"line-height: 1.4\">I can see the benefit of working in a team now - much faster generation of ideas.&nbsp;</span></p>\n<p>However, I think I'm done now - I don't have anymore time to make changes before the deadline :(</p>\n<p><span style=\"line-height: 1.4\">Definitely looked forward to reading about the winners models.</span></p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "28811",
      "postDate": "08/16/2013 11:22:01",
      "content": "<p>Hi, EL, I think the winning tips might be that 'how could you learn from this seemingly impossible scatter data ?' I am also stucked here at private test CV score 0.75.&nbsp;</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "28827",
      "postDate": "08/16/2013 17:45:14",
      "content": "<p>Hi</p>\n<p>it is often very slow to upload the model via kaggle website.</p>\n<p>Do you have any alternative way?</p>\n<p>Thanks</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "28828",
      "postDate": "08/16/2013 17:48:23",
      "content": "<p>[quote=LucaToni;28827]</p>\n<p>Hi</p>\n<p>it is often very slow to upload the model via kaggle website.</p>\n<p>Do you have any alternative way?</p>\n<p>Thanks</p>\n<p>[/quote]</p>\n<p>&nbsp;</p>\n<p>works fine with me. what is the total size of your model? mine is around 12mb</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "28904",
      "postDate": "08/18/2013 18:04:59",
      "content": "<p>REMINDER:</p>\n<p>You zip file and directory when you submit your code have to bear the name of your team and you need to provide a contact name and email in the README file.</p>\n<p>The following submissions are corrupted:</p>\n<p>Saeh &amp; Xing (connot retrive, please provide a backup way of downloading)</p>\n<p>sbachish (predictions were submitted instead of model)</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "28905",
      "postDate": "08/18/2013 18:42:23",
      "content": "<p>I've uploaded another model with a new readme file</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "28927",
      "postDate": "08/19/2013 01:02:13",
      "content": "<p>I succeeded in downloading&nbsp;Saeh &amp; Xing with the help of Kaggle support.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "28988",
      "postDate": "08/19/2013 22:09:44",
      "content": "<p>Hi Isabelle,</p>\n<p>I think it was because i used an ampersand (&amp;) in the file name as i too couldn't download the file even directly after i uploaded it.</p>\n<p>&nbsp;</p>\n<p>I've fixed the filename issue now and the download should work. The new file has exactly the same contents as the old one.</p>\n<p>Thanks</p>",
      "rawMarkdown": "",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 28654,
      "author_name": "iguyon",
      "author_url": "",
      "post_date": "08/14/2013 00:37:24",
      "content": "<p>Thank you very much to those who started uploading their code. The others should REALLY do this as soon as possible. Remember:</p>\n<p>- Name the archive after your team name</p>\n<p>- Include ONE up-to-date README file with</p>\n<p style=\"padding-left: 30px\">* You team name</p>\n<p style=\"padding-left: 30px\">* A contact EMAIL</p>\n<p style=\"padding-left: 30px\">* SIMPLE and complete instructions (including ALL the software dependencies)</p>\n<p>- Do NOT include a copy of the whole dataset as part of your submission. Add a SMALL data sample and a SMALL test (running is a few seconds) so the code can be checked rapidly</p>\n<p>- If training software is included (optional) include instructions to process the final test data WITHOUT retraining (i.e. include a TRAINED MODEL for which predict can just be run without retraining.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 28656,
      "author_name": "liubenyuan",
      "author_url": "",
      "post_date": "08/14/2013 02:06:48",
      "content": "<p>can I assume that all the csv data files are available by default ? and is the .mat file valid for the submission purpose ?&nbsp;</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 28657,
      "author_name": "iguyon",
      "author_url": "",
      "post_date": "08/14/2013 02:17:27",
      "content": "<p>The final test data files will be called&nbsp;CEfinal_test_pairs.csv and&nbsp;CEfinal_test_publicinfo.csv. For the split format, there will be a file&nbsp;CEfinal_test_split, with files test1.txt, test2.txr, etc.</p>\n<p>You don't need to attach data to your code (unless you want to provide a small sample to test purposes).</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 28677,
      "author_name": "bram278806",
      "author_url": "",
      "post_date": "08/14/2013 11:32:40",
      "content": "<p>Hello,</p>\n<p><span style=\"line-height: 1.4\">Can we assume that your Python environment already contains the following typical libraries?</span></p>\n<ul>\n<li>numpy</li>\n<li>pandas</li>\n<li>scipy</li>\n<li>sklearn</li>\n<li>pylab</li>\n</ul>\n<p>Thank you.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 28700,
      "author_name": "iguyon",
      "author_url": "",
      "post_date": "08/14/2013 17:06:26",
      "content": "<p>You can indicate which library you use and the version and we can try to install the libraries and check whether everything works fine. But this is at your own risks. I recommend that you provide a piece of test code we can run that checks the full installation by training and testing on a small data subset and checks the values obtained are identical to those produced when you run the code on your computer. In particular, make sure that you do not depend on randomly generated numbers that change your result. Draw random numbers in advance and always use the same values.</p>\n<p>Note: for verification, we do not require retraining, so consider re-loading a trained model and using it to make predictions on the test data.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 28728,
      "author_name": "jarjestelmacom",
      "author_url": "",
      "post_date": "08/15/2013 12:47:41",
      "content": "<p>Is it ok for the program to print some statistics or timing information that doesn't stay the same every run? I.e. how many seconds/milliseconds it took to read a file, do analysis 1, do analysis 2 etc..</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 28732,
      "author_name": "iguyon",
      "author_url": "",
      "post_date": "08/15/2013 13:34:27",
      "content": "<p>Yes. It is actually a good idea to print some diagnostics.&nbsp;</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 28734,
      "author_name": "liubenyuan",
      "author_url": "",
      "post_date": "08/15/2013 13:38:05",
      "content": "<p>I think I won't be able to improve much above 0.75, I tried to dig more features out of the scatter data but currently I am using only one RF regressor. Could any one hint some light on improving further the results using some more **advanced** regressors ?&nbsp;</p>\n<p>And I am very keen to see the method that was used to reach Rank1!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 28736,
      "author_name": "abhishek",
      "author_url": "",
      "post_date": "08/15/2013 14:26:37",
      "content": "<p>[quote=liubenyuan;28734]</p>\n<p>I think I won't be able to improve much above 0.75, I tried to dig more features out of the scatter data but currently I am using only one RF regressor. Could any one hint some light on improving further the results using some more **advanced** regressors ?&nbsp;</p>\n<p>And I am very keen to see the method that was used to reach Rank1!</p>\n<p>[/quote]</p>\n<p>&nbsp;</p>\n<p>did you also use the supplimentary data?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 28738,
      "author_name": "liubenyuan",
      "author_url": "",
      "post_date": "08/15/2013 14:54:16",
      "content": "<p>Yes, both SUP1 and SUP2 are used.&nbsp;</p>\n<p>I found that the CV test on SUP2 provided a good match to the final valid dataset.&nbsp;</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 28739,
      "author_name": "abhishek",
      "author_url": "",
      "post_date": "08/15/2013 14:56:54",
      "content": "<p>[quote=liubenyuan;28738]</p>\n<p>Yes, both SUP1 and SUP2 are used.&nbsp;</p>\n<p>I found that the CV test on SUP2 provided a good match to the final valid dataset.&nbsp;</p>\n<p>[/quote]</p>\n<p>&nbsp;</p>\n<p>what about SUP3? any specific reason for not using it? would you mind telling on how many features you are training your model?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 28740,
      "author_name": "liubenyuan",
      "author_url": "",
      "post_date": "08/15/2013 15:02:35",
      "content": "<p>81 features, most of them are from the python code. and to those who use matlab :</p>\n<p>1. ANM, ANMD, PNL are far too slow to be used to extract the features. and ANM and ANMD reported NaN results on some pairs (warning).</p>\n<p>2. LINGAM is fast, however the result on the quantized data are not so good.</p>\n<p>3. the nips-code folder in GPI is very valuable, and I think the features extracted using this method might be good for training. However they are relatively too slow running in my laptop ( ~approximately 7 days on the whole train, SUP1, SUP2 datasets). some one may give it a try !!</p>\n<p>4. IGCI is extremely fast, some of my other features are generated using variant versions of this code.&nbsp;</p>\n<p>5. and some fabricated features by intuition of viewing (ax,bx) paris.</p>\n<p>&nbsp;</p>\n<p>It is a great challenge. I hope these information might help in these days.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 28751,
      "author_name": "eoinlawless",
      "author_url": "",
      "post_date": "08/15/2013 18:26:25",
      "content": "<p>I also had good success with the Lingam algorithms. I have only 27 features left after I did a round of feature selection recently. I tried additive noise models, but could not get good results from them. I tried training on each target value one at a time and then using the results of that as features for a second round of testing, but again the results were worse if anything. &nbsp;I found IGCI worked well on the initial data but not so well on the 'fixed' data.</p>\n<p>&nbsp;</p>\n<p>By some last minute blending of models I've improved my score by 0.01, but I suspect to go further I'd need to create some new features (ie get PNL, ANM or something new working).</p>\n<p>&nbsp;</p>\n<p>My 10-fold CV score is actually 0.76 (compared to 0.72 on the Leaderboard). How do other people's CV scores compare to the leaderboard?&nbsp;</p>\n<p>I've learnt an incredible amount from this competition - thank you Isabelle and Kaggle.</p>\n<p>&nbsp;</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 28766,
      "author_name": "kinnskogr",
      "author_url": "",
      "post_date": "08/15/2013 21:04:48",
      "content": "<p>Hi Eoin,</p>\n<p>&nbsp; &nbsp;My CV and leaderboard scores have actually been quite consistent. For my latest submissions, I got&nbsp;0.701680 CV vs&nbsp;0.71140 leaderboard. Others have been similarly close, usually within the RMS of the CV scores.</p>\n<p>&nbsp; &nbsp;At one point, my classifier was somewhat asymmetric, performing better for backward AUC than for forward AUC. CV scores on those models tended to have a larger deviance with respect to the leaderboard scores, possibly because of differences in sample composition. Are your models all close to symmetric?</p>\n<p>&nbsp;</p>\n<p>- Emanuel</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 28768,
      "author_name": "eoinlawless",
      "author_url": "",
      "post_date": "08/15/2013 21:15:07",
      "content": "<p>Hi Emmanuel,</p>\n<p>I think you're right there - my models are giving asymmetric scores, with a difference of almost 0.1. Interesting... probably too late to change my code, but not too late to retrain ;)</p>\n<p>&nbsp;</p>\n<p>Eoin</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 28781,
      "author_name": "kinnskogr",
      "author_url": "",
      "post_date": "08/16/2013 01:24:39",
      "content": "<p>Well, I wish you good luck (but not too much luck ;) )&nbsp;</p>\n<p>&nbsp;</p>\n<p>- Emanuel</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 28798,
      "author_name": "eoinlawless",
      "author_url": "",
      "post_date": "08/16/2013 07:47:33",
      "content": "<p>I fixed an asymmetry in my model and my score improved slightly. Still searching for the magic sauce that the people above 0.8 are using...</p>\n<p>&nbsp;</p>\n<p>Eoin</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 28799,
      "author_name": "abhishek",
      "author_url": "",
      "post_date": "08/16/2013 07:50:52",
      "content": "<p>[quote=Eoin Lawless;28798]</p>\n<p>I fixed an asymmetry in my model and my score improved slightly. Still searching for the magic sauce that the people above 0.8 are using...</p>\n<p>&nbsp;</p>\n<p>Eoin</p>\n<p>[/quote]</p>\n<p>&nbsp;</p>\n<p>damn, u took our position..</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 28800,
      "author_name": "eoinlawless",
      "author_url": "",
      "post_date": "08/16/2013 08:53:54",
      "content": "<p>I'm sure there'll be a lot of activity over the next few days... I don't expect to hold it for long...</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 28801,
      "author_name": "abhishek",
      "author_url": "",
      "post_date": "08/16/2013 08:55:36",
      "content": "<p>[quote=Eoin Lawless;28800]</p>\n<p>I'm sure there'll be a lot of activity over the next few days... I don't expect to hold it for long...</p>\n<p>[/quote]</p>\n<p>&nbsp;</p>\n<p>yeah, seems so. new teams coming up and huge increase in ranks and all...</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 28804,
      "author_name": "liubenyuan",
      "author_url": "",
      "post_date": "08/16/2013 09:21:32",
      "content": "<p>yeah, it was also exciting to know handling the asymmetric could improve another 0.02 :P</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 28806,
      "author_name": "eoinlawless",
      "author_url": "",
      "post_date": "08/16/2013 09:34:42",
      "content": "<p>I only got 0.011 from the fix, but my CV scores are still higher than the leaderboard score so I may still be missing something.&nbsp;</p>\n<p><span style=\"line-height: 1.4\">I can see the benefit of working in a team now - much faster generation of ideas.&nbsp;</span></p>\n<p>However, I think I'm done now - I don't have anymore time to make changes before the deadline :(</p>\n<p><span style=\"line-height: 1.4\">Definitely looked forward to reading about the winners models.</span></p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 28811,
      "author_name": "liubenyuan",
      "author_url": "",
      "post_date": "08/16/2013 11:22:01",
      "content": "<p>Hi, EL, I think the winning tips might be that 'how could you learn from this seemingly impossible scatter data ?' I am also stucked here at private test CV score 0.75.&nbsp;</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 28827,
      "author_name": "gbonte",
      "author_url": "",
      "post_date": "08/16/2013 17:45:14",
      "content": "<p>Hi</p>\n<p>it is often very slow to upload the model via kaggle website.</p>\n<p>Do you have any alternative way?</p>\n<p>Thanks</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 28828,
      "author_name": "abhishek",
      "author_url": "",
      "post_date": "08/16/2013 17:48:23",
      "content": "<p>[quote=LucaToni;28827]</p>\n<p>Hi</p>\n<p>it is often very slow to upload the model via kaggle website.</p>\n<p>Do you have any alternative way?</p>\n<p>Thanks</p>\n<p>[/quote]</p>\n<p>&nbsp;</p>\n<p>works fine with me. what is the total size of your model? mine is around 12mb</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 28904,
      "author_name": "iguyon",
      "author_url": "",
      "post_date": "08/18/2013 18:04:59",
      "content": "<p>REMINDER:</p>\n<p>You zip file and directory when you submit your code have to bear the name of your team and you need to provide a contact name and email in the README file.</p>\n<p>The following submissions are corrupted:</p>\n<p>Saeh &amp; Xing (connot retrive, please provide a backup way of downloading)</p>\n<p>sbachish (predictions were submitted instead of model)</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 28905,
      "author_name": "domcastro",
      "author_url": "",
      "post_date": "08/18/2013 18:42:23",
      "content": "<p>I've uploaded another model with a new readme file</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 28927,
      "author_name": "iguyon",
      "author_url": "",
      "post_date": "08/19/2013 01:02:13",
      "content": "<p>I succeeded in downloading&nbsp;Saeh &amp; Xing with the help of Kaggle support.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 28988,
      "author_name": "thomasatkins",
      "author_url": "",
      "post_date": "08/19/2013 22:09:44",
      "content": "<p>Hi Isabelle,</p>\n<p>I think it was because i used an ampersand (&amp;) in the file name as i too couldn't download the file even directly after i uploaded it.</p>\n<p>&nbsp;</p>\n<p>I've fixed the filename issue now and the download should work. The new file has exactly the same contents as the old one.</p>\n<p>Thanks</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "28184": "",
    "28654": "",
    "28656": "",
    "28657": "",
    "28677": "",
    "28700": "",
    "28728": "",
    "28732": "",
    "28734": "",
    "28736": "",
    "28738": "",
    "28739": "",
    "28740": "",
    "28751": "",
    "28766": "",
    "28768": "",
    "28781": "",
    "28798": "",
    "28799": "",
    "28800": "",
    "28801": "",
    "28804": "",
    "28806": "",
    "28811": "",
    "28827": "",
    "28828": "",
    "28904": "",
    "28905": "",
    "28927": "",
    "28988": ""
  },
  "source": "meta"
}