{ "id": 266230, "title": "Data Augmentation", "url": "/competitions/g2net-gravitational-wave-detection/discussion/266230", "author_name": "", "post_date": "2021-08-18T10:49:51.029743600Z", "votes": 16, "comment_count": 13, "views": 0, "content": "

To reduce the risk of overfitting,i simply added two kinds of data augmentations:
\n1.shuffle the three files order before CQT trans;
\n2.add Cutout data augmentation
\nThese can help me improve my cv 0.8521=>0.8553,and LB 0.867=>0.868

", "messages": [ { "id": "1479213", "postDate": "08/18/2021 10:49:51", "content": "

To reduce the risk of overfitting,i simply added two kinds of data augmentations:
\n1.shuffle the three files order before CQT trans;
\n2.add Cutout data augmentation
\nThese can help me improve my cv 0.8521=>0.8553,and LB 0.867=>0.868

", "rawMarkdown": "To reduce the risk of overfitting,i simply added two kinds of data augmentations:\n1.shuffle the three files order before CQT trans;\n2.add Cutout data augmentation\nThese can help me improve my cv 0.8521=>0.8553,and LB 0.867=>0.868", "votes": null }, { "id": "1479253", "postDate": "08/18/2021 11:13:53", "content": "

Hi @hanson0910 you always apply the first augmentation (shuffle the three files order before CQT trans), or with a probability?. Thanks

", "rawMarkdown": "Hi @hanson0910 you always apply the first augmentation (shuffle the three files order before CQT trans), or with a probability?. Thanks", "votes": null }, { "id": "1479277", "postDate": "08/18/2021 11:31:22", "content": "

with p=0.5

", "rawMarkdown": "with p=0.5", "votes": null }, { "id": "1481538", "postDate": "08/19/2021 15:12:39", "content": "

Did you try mixup? I used mixup and it does not seem to work well due to noise in the data.

", "rawMarkdown": "Did you try mixup? I used mixup and it does not seem to work well due to noise in the data.", "votes": null }, { "id": "1481546", "postDate": "08/19/2021 15:16:58", "content": "

i didn't use mixup,i think mixup may break the time information.

", "rawMarkdown": "i didn't use mixup,i think mixup may break the time information.", "votes": null }, { "id": "1481550", "postDate": "08/19/2021 15:19:59", "content": "

I used mixup augmentation on the 2D spectrogram generated by CQT trans. It's reported in the SETI competition that OR Mixup improved the model's performance.

", "rawMarkdown": "I used mixup augmentation on the 2D spectrogram generated by CQT trans. It's reported in the SETI competition that OR Mixup improved the model's performance.", "votes": null }, { "id": "1481567", "postDate": "08/19/2021 15:26:00", "content": "

thanks for your share,i will try it.

", "rawMarkdown": "thanks for your share,i will try it.", "votes": null }, { "id": "1485144", "postDate": "08/21/2021 20:41:07", "content": "

Do you mean you shuffle signals from three detectors before stacking? It's one file, not three files.

", "rawMarkdown": "Do you mean you shuffle signals from three detectors before stacking? It's one file, not three files.", "votes": null }, { "id": "1486472", "postDate": "08/23/2021 02:22:08", "content": "

you could understand one file with 3 channels

", "rawMarkdown": "you could understand one file with 3 channels", "votes": null }, { "id": "1486816", "postDate": "08/23/2021 08:28:34", "content": "

@solosquad1999 I also tried mixup and didn't get any significant improvement, did it work for you later ?

", "rawMarkdown": "solosquad1999 I also tried mixup and didn't get any significant improvement, did it work for you later ?", "votes": null }, { "id": "1486894", "postDate": "08/23/2021 09:32:10", "content": "

@zarif98sjs : Neither does my model.

", "rawMarkdown": "zarif98sjs : Neither does my model.", "votes": null }, { "id": "1486978", "postDate": "08/23/2021 10:25:21", "content": "

The same for me :( Mixup does not work.

", "rawMarkdown": "The same for me :( Mixup does not work.", "votes": null }, { "id": "1487287", "postDate": "08/23/2021 14:23:26", "content": "

I wonder why it worked so good for the SETI competition : ( (I didn't give that, so don't know much about the data)

", "rawMarkdown": "I wonder why it worked so good for the SETI competition : ( (I didn't give that, so don't know much about the data)", "votes": null }, { "id": "1561184", "postDate": "10/27/2021 12:15:21", "content": "

Hey All,

\n

Thank you all for taking part in our competition. The participation has been overwhelmingly positive. We are currently conducting a survey to gauge the demographic and outreach achieved. Kindly spare 2min and fill in this survey https://forms.gle/QP9L16niPexozyhu5.

\n

Thank you all,

\n

Regards,
\nChris

", "rawMarkdown": "Hey All,\n\nThank you all for taking part in our competition. The participation has been overwhelmingly positive. We are currently conducting a survey to gauge the demographic and outreach achieved. Kindly spare 2min and fill in this survey https://forms.gle/QP9L16niPexozyhu5.\n\nThank you all,\n\nRegards,\nChris", "votes": null } ], "comments": [ { "id": 1479253, "author_name": "lhkhiem28", "author_url": "", "post_date": "08/18/2021 11:13:53", "content": "

Hi @hanson0910 you always apply the first augmentation (shuffle the three files order before CQT trans), or with a probability?. Thanks

", "votes": null, "replies": [ { "id": 1479277, "author_name": "hanson0910", "author_url": "", "post_date": "08/18/2021 11:31:22", "content": "

with p=0.5

", "votes": null, "replies": [] } ] }, { "id": 1481538, "author_name": "solosquad1999", "author_url": "", "post_date": "08/19/2021 15:12:39", "content": "

Did you try mixup? I used mixup and it does not seem to work well due to noise in the data.

", "votes": null, "replies": [ { "id": 1481546, "author_name": "hanson0910", "author_url": "", "post_date": "08/19/2021 15:16:58", "content": "

i didn't use mixup,i think mixup may break the time information.

", "votes": null, "replies": [] }, { "id": 1481550, "author_name": "solosquad1999", "author_url": "", "post_date": "08/19/2021 15:19:59", "content": "

I used mixup augmentation on the 2D spectrogram generated by CQT trans. It's reported in the SETI competition that OR Mixup improved the model's performance.

", "votes": null, "replies": [] }, { "id": 1481567, "author_name": "hanson0910", "author_url": "", "post_date": "08/19/2021 15:26:00", "content": "

thanks for your share,i will try it.

", "votes": null, "replies": [] }, { "id": 1486816, "author_name": "zarif98sjs", "author_url": "", "post_date": "08/23/2021 08:28:34", "content": "

@solosquad1999 I also tried mixup and didn't get any significant improvement, did it work for you later ?

", "votes": null, "replies": [] }, { "id": 1486894, "author_name": "solosquad1999", "author_url": "", "post_date": "08/23/2021 09:32:10", "content": "

@zarif98sjs : Neither does my model.

", "votes": null, "replies": [] }, { "id": 1486978, "author_name": "analokamus", "author_url": "", "post_date": "08/23/2021 10:25:21", "content": "

The same for me :( Mixup does not work.

", "votes": null, "replies": [] }, { "id": 1487287, "author_name": "zarif98sjs", "author_url": "", "post_date": "08/23/2021 14:23:26", "content": "

I wonder why it worked so good for the SETI competition : ( (I didn't give that, so don't know much about the data)

", "votes": null, "replies": [] } ] }, { "id": 1485144, "author_name": "atamazian", "author_url": "", "post_date": "08/21/2021 20:41:07", "content": "

Do you mean you shuffle signals from three detectors before stacking? It's one file, not three files.

", "votes": null, "replies": [ { "id": 1486472, "author_name": "hanson0910", "author_url": "", "post_date": "08/23/2021 02:22:08", "content": "

you could understand one file with 3 channels

", "votes": null, "replies": [] } ] }, { "id": 1561184, "author_name": "zerafachris", "author_url": "", "post_date": "10/27/2021 12:15:21", "content": "

Hey All,

\n

Thank you all for taking part in our competition. The participation has been overwhelmingly positive. We are currently conducting a survey to gauge the demographic and outreach achieved. Kindly spare 2min and fill in this survey https://forms.gle/QP9L16niPexozyhu5.

\n

Thank you all,

\n

Regards,
\nChris

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1479213": "To reduce the risk of overfitting,i simply added two kinds of data augmentations:\n1.shuffle the three files order before CQT trans;\n2.add Cutout data augmentation\nThese can help me improve my cv 0.8521=>0.8553,and LB 0.867=>0.868", "1479253": "Hi @hanson0910 you always apply the first augmentation (shuffle the three files order before CQT trans), or with a probability?. Thanks", "1479277": "with p=0.5", "1481538": "Did you try mixup? I used mixup and it does not seem to work well due to noise in the data.", "1481546": "i didn't use mixup,i think mixup may break the time information.", "1481550": "I used mixup augmentation on the 2D spectrogram generated by CQT trans. It's reported in the SETI competition that OR Mixup improved the model's performance.", "1481567": "thanks for your share,i will try it.", "1485144": "Do you mean you shuffle signals from three detectors before stacking? It's one file, not three files.", "1486472": "you could understand one file with 3 channels", "1486816": "solosquad1999 I also tried mixup and didn't get any significant improvement, did it work for you later ?", "1486894": "zarif98sjs : Neither does my model.", "1486978": "The same for me :( Mixup does not work.", "1487287": "I wonder why it worked so good for the SETI competition : ( (I didn't give that, so don't know much about the data)", "1561184": "Hey All,\n\nThank you all for taking part in our competition. The participation has been overwhelmingly positive. We are currently conducting a survey to gauge the demographic and outreach achieved. Kindly spare 2min and fill in this survey https://forms.gle/QP9L16niPexozyhu5.\n\nThank you all,\n\nRegards,\nChris" }, "source": "meta" }