We've just kicked off this short competition around semi-supervised feature learning, and hope that it will provide an interesting task and a good range of comparisons. We're particularly hopeful that there will be a wide diversity of methods applied.
\r\nPlease feel free to share thoughts and ideas on this forum, and also to post questions if you need help with any of the data.
\r\nGood luck, and thanks for taking part!
", "messages": [ { "id": "5645", "postDate": "09/25/2011 20:14:52", "content": "We've just kicked off this short competition around semi-supervised feature learning, and hope that it will provide an interesting task and a good range of comparisons. We're particularly hopeful that there will be a wide diversity of methods applied.
\r\nPlease feel free to share thoughts and ideas on this forum, and also to post questions if you need help with any of the data.
\r\nGood luck, and thanks for taking part!
", "rawMarkdown": "", "votes": null }, { "id": "5658", "postDate": "09/27/2011 04:29:48", "content": "I can't seem to view unlabeled_data.svmlight.dat. Does that format mean support vector machines? How would I view it, I have tried notepad++ and excel. There seems to be too much data.\r\n
\r\nThanks
Hi Aniket,
\r\nThis data file is a plain ASCII text file -- you should be able to view it in a text editor if nothing else.
\r\nThe first line looks like this:
\r\n0 13550:1 20936:1 26593:1 36741:1 40526:1 53673:1 55629:1 58808:1 61631:1 75800:1 91692:0.142857 107242:1 112394:1 113575:1 118203:1 126690:1 132129:1 133569:1 143017:1 156931:1 158844:1 174906:1 178590:1 181155:1 202154:1 206500:1 230427:1 240222:1 264924:1\r\n 286247:1 320279:1 334469:1 338381:1 341804:1 346019:1 360928:1 363210:1 366389:0.033195 400466:1 406025:1 416076:1 418463:1 418544:1 422442:1 423245:1 427521:1 465289:1 471149:1 492910:1 493742:1 511517:1 514178:1 519958:1 520398:1 545486:1 547291:1 551357:1\r\n 583080:1 598032:1 599175:1 600788:1 609497:1 611335:1 630305:0.00595074 634156:1 656292:1 680891:1 685138:1 689077:1 702637:1 717017:1 719516:0.914598 719963:1 726524:1 729005:1 759448:1 771272:1 780398:1 810120:1 814090:1 825809:1 826357:1 831823:1 855818:1\r\n 860370:1 861577:1 870014:1 877134:0.0551724 877813:1 908872:1 910572:0.2002 921965:0.0588235 922498:1 928586:0.00595238 931054:1 934933:1 937057:1 938549:1 942980:1 943012:1 949128:1 954763:0.00595238 957637:1 974200:1
\r\nThe first entry "0" shows that the example is not labeled. The next entry "13550:1 " shows that the feature with id number 13550 has a value of 1. The rest of the entries show other non-zero feature values, on up through feature id 974200, which has a\r\n value of 1. Features that are not listed are implied to have a value of 0.
\r\nThis is the data format used by SVM-light, libsvm, and several other suppor vector machine packages. We chose it because it compactly represents sparse, high dimensional data.
\r\nHope this is helpful. Let us know if you have any other questions about it.
", "rawMarkdown": "", "votes": null }, { "id": "5666", "postDate": "09/27/2011 16:40:01", "content": "Still can't open this file. It says the file is too big. Size: 1,079,505 KB. Any suggestions what editor to try? I have tried notepad, notepad++, and excel.
", "rawMarkdown": "", "votes": null }, { "id": "5669", "postDate": "09/27/2011 17:00:55", "content": "Try Ultraedit.
", "rawMarkdown": "", "votes": null }, { "id": "5670", "postDate": "09/27/2011 17:01:27", "content": "You can extract the first 10 lines using this Python script or use the more command in unix.
\r\n\r\n f1=open('unlabeled_data.svmlight.dat','r')
f2=open('unlabeled_data.svmlight2.dat','w')
i=0
for line in f1:
if i f2.write(line)
i+=1
else:
break
f1.close()
f2.close()
\r\n",
"rawMarkdown": "",
"votes": null
},
{
"id": "5676",
"postDate": "09/27/2011 18:03:11",
"content": "apparently vim can open it, pretty sure it is showing me all the data.
", "rawMarkdown": "", "votes": null }, { "id": "5687", "postDate": "09/27/2011 21:34:37", "content": "Would someone please describe in detail (or point me to a detailed description) of the classifier that will be used and how it will be trained. Sorry, if it's here somewhere and I just missed it. I looked.
", "rawMarkdown": "", "votes": null }, { "id": "5688", "postDate": "09/27/2011 21:47:51", "content": "The columns in the example_submission_file are the features? So there are 100 features plus the first column is a classification of either 0 or 1?
\r\n\r\n
Thanks
", "rawMarkdown": "", "votes": null } ], "comments": [ { "id": 5658, "author_name": "", "author_url": "", "post_date": "09/27/2011 04:29:48", "content": "I can't seem to view unlabeled_data.svmlight.dat. Does that format mean support vector machines? How would I view it, I have tried notepad++ and excel. There seems to be too much data.\r\n
\r\nThanks
Hi Aniket,
\r\nThis data file is a plain ASCII text file -- you should be able to view it in a text editor if nothing else.
\r\nThe first line looks like this:
\r\n0 13550:1 20936:1 26593:1 36741:1 40526:1 53673:1 55629:1 58808:1 61631:1 75800:1 91692:0.142857 107242:1 112394:1 113575:1 118203:1 126690:1 132129:1 133569:1 143017:1 156931:1 158844:1 174906:1 178590:1 181155:1 202154:1 206500:1 230427:1 240222:1 264924:1\r\n 286247:1 320279:1 334469:1 338381:1 341804:1 346019:1 360928:1 363210:1 366389:0.033195 400466:1 406025:1 416076:1 418463:1 418544:1 422442:1 423245:1 427521:1 465289:1 471149:1 492910:1 493742:1 511517:1 514178:1 519958:1 520398:1 545486:1 547291:1 551357:1\r\n 583080:1 598032:1 599175:1 600788:1 609497:1 611335:1 630305:0.00595074 634156:1 656292:1 680891:1 685138:1 689077:1 702637:1 717017:1 719516:0.914598 719963:1 726524:1 729005:1 759448:1 771272:1 780398:1 810120:1 814090:1 825809:1 826357:1 831823:1 855818:1\r\n 860370:1 861577:1 870014:1 877134:0.0551724 877813:1 908872:1 910572:0.2002 921965:0.0588235 922498:1 928586:0.00595238 931054:1 934933:1 937057:1 938549:1 942980:1 943012:1 949128:1 954763:0.00595238 957637:1 974200:1
\r\nThe first entry "0" shows that the example is not labeled. The next entry "13550:1 " shows that the feature with id number 13550 has a value of 1. The rest of the entries show other non-zero feature values, on up through feature id 974200, which has a\r\n value of 1. Features that are not listed are implied to have a value of 0.
\r\nThis is the data format used by SVM-light, libsvm, and several other suppor vector machine packages. We chose it because it compactly represents sparse, high dimensional data.
\r\nHope this is helpful. Let us know if you have any other questions about it.
", "votes": null, "replies": [] }, { "id": 5666, "author_name": "", "author_url": "", "post_date": "09/27/2011 16:40:01", "content": "Still can't open this file. It says the file is too big. Size: 1,079,505 KB. Any suggestions what editor to try? I have tried notepad, notepad++, and excel.
", "votes": null, "replies": [] }, { "id": 5669, "author_name": "", "author_url": "", "post_date": "09/27/2011 17:00:55", "content": "Try Ultraedit.
", "votes": null, "replies": [] }, { "id": 5670, "author_name": "dirknbr", "author_url": "", "post_date": "09/27/2011 17:01:27", "content": "You can extract the first 10 lines using this Python script or use the more command in unix.
\r\n\r\n f1=open('unlabeled_data.svmlight.dat','r')
f2=open('unlabeled_data.svmlight2.dat','w')
i=0
for line in f1:
if i f2.write(line)
i+=1
else:
break
f1.close()
f2.close()
\r\n",
"votes": null,
"replies": []
},
{
"id": 5676,
"author_name": "",
"author_url": "",
"post_date": "09/27/2011 18:03:11",
"content": "apparently vim can open it, pretty sure it is showing me all the data.
", "votes": null, "replies": [] }, { "id": 5687, "author_name": "byrond", "author_url": "", "post_date": "09/27/2011 21:34:37", "content": "Would someone please describe in detail (or point me to a detailed description) of the classifier that will be used and how it will be trained. Sorry, if it's here somewhere and I just missed it. I looked.
", "votes": null, "replies": [] }, { "id": 5688, "author_name": "", "author_url": "", "post_date": "09/27/2011 21:47:51", "content": "The columns in the example_submission_file are the features? So there are 100 features plus the first column is a classification of either 0 or 1?
\r\n\r\n
Thanks
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "5645": "", "5658": "", "5660": "", "5666": "", "5669": "", "5670": "", "5676": "", "5687": "", "5688": "" }, "source": "meta" }