As the competition nears completion, we thought it would be good to provide a little more detail about the final submission process.
\r\nWe've enabled each team to select four of their submissions for finale evaluation. You should definitely pick your best performing submission as one of these four. For the others, feel free to pick submissions that use different methods or approaches.\r\n Because this is a research oriented competition, it will make the results more interesting to know about which methods that don't work as well in addition to methods that do work well.
\r\nRemember that, as described on the Evaluation page, part of the final submission process involves sending a short description of each of your final submission methods to semisupervisedfeatures@gmail.com no later than 11:59 PM (UTC) on Monday, Oct. 17th.\r\n This must be done to be eligible for the prize, or for your results to be included in the final paper for the results of this competition. If you use a method that is previously published, please include a citation. (BibTex format references are appreciated\r\n but not required.)
\r\nHere is an example writeup, for the "Benchmark: minibatch k-means" result:
\r\nThis submission used mini-batch k-means [1] to learn the cluster centers. We trained 100 cluster centers on a concatenated data set using the 1M unlabled examples and the 50k labeled training examples. We used the mini-batch k-means learning from sofia-ml\r\n [2], with mini-batches of size 1000 and 10000 iterations. Training the cluster centers took 25 minutes on a 2GHz laptop. Each new example was mapped to a space of 100 features, one per cluster center, using the RBF kernel value between the example and the\r\n given cluster center. The RBF kernel parameter was 0.01.
\r\nReferences:
\r\n[1] http://code.google.com/p/sofia-ml/
\r\n[2] @inproceedings{minibatchKmeans2010,
\r\n title={Web-scale k-means clustering},\r\n author={Sculley, D.},\r\n booktitle={Proceedings of the 19th international conference on World wide web},\r\n pages={1177--1178},\r\n year={2010},\r\n organization={ACM}\r\n}",
"messages": [
{
"id": 6092,
"postDate": "2011-10-16T21:30:37.400Z",
"content": "As the competition nears completion, we thought it would be good to provide a little more detail about the final submission process.
\r\nWe've enabled each team to select four of their submissions for finale evaluation. You should definitely pick your best performing submission as one of these four. For the others, feel free to pick submissions that use different methods or approaches.\r\n Because this is a research oriented competition, it will make the results more interesting to know about which methods that don't work as well in addition to methods that do work well.
\r\nRemember that, as described on the Evaluation page, part of the final submission process involves sending a short description of each of your final submission methods to semisupervisedfeatures@gmail.com no later than 11:59 PM (UTC) on Monday, Oct. 17th.\r\n This must be done to be eligible for the prize, or for your results to be included in the final paper for the results of this competition. If you use a method that is previously published, please include a citation. (BibTex format references are appreciated\r\n but not required.)
\r\nHere is an example writeup, for the "Benchmark: minibatch k-means" result:
\r\nThis submission used mini-batch k-means [1] to learn the cluster centers. We trained 100 cluster centers on a concatenated data set using the 1M unlabled examples and the 50k labeled training examples. We used the mini-batch k-means learning from sofia-ml\r\n [2], with mini-batches of size 1000 and 10000 iterations. Training the cluster centers took 25 minutes on a 2GHz laptop. Each new example was mapped to a space of 100 features, one per cluster center, using the RBF kernel value between the example and the\r\n given cluster center. The RBF kernel parameter was 0.01.
\r\nReferences:
\r\n[1] http://code.google.com/p/sofia-ml/
\r\n[2] @inproceedings{minibatchKmeans2010,
\r\n title={Web-scale k-means clustering},\r\n author={Sculley, D.},\r\n booktitle={Proceedings of the 19th international conference on World wide web},\r\n pages={1177--1178},\r\n year={2010},\r\n organization={ACM}\r\n}"
}
],
"comments": [],
"raw_markdown_by_id": {
"6092": ""
}
}