{ "id": 119143, "title": "Time spent so far", "url": "/competitions/tensorflow2-question-answering/discussion/119143", "author_name": "", "post_date": "2019-11-26T21:03:39.736250700Z", "votes": 22, "comment_count": 5, "views": 0, "content": "

\"\"

", "messages": [ { "id": "682047", "postDate": "11/26/2019 21:03:39", "content": "

\"\"

", "rawMarkdown": "![](https://i.imgflip.com/3hn908.jpg)", "votes": null }, { "id": "682287", "postDate": "11/27/2019 07:27:04", "content": "

Nice summary of all recent competitions with a custom metric on Kaggle. Maybe you could also add \"waiting for metric to be fixed\" somewhere :)

\n\n

Seriously, it would be great if all competitions with a custom metric had actual metric code published (not just the inner loop like in https://www.kaggle.com/c/pku-autonomous-driving/overview/evaluation). Even better if it's possible to move to python metric implementation, this may also reduce amount of bugs or time to fix them, because participants will be able to help, and you could use existing metric implementations more.

", "rawMarkdown": "Nice summary of all recent competitions with a custom metric on Kaggle. Maybe you could also add \"waiting for metric to be fixed\" somewhere :)\n\nSeriously, it would be great if all competitions with a custom metric had actual metric code published (not just the inner loop like in https://www.kaggle.com/c/pku-autonomous-driving/overview/evaluation). Even better if it's possible to move to python metric implementation, this may also reduce amount of bugs or time to fix them, because participants will be able to help, and you could use existing metric implementations more.", "votes": null }, { "id": "682298", "postDate": "11/27/2019 08:06:02", "content": "

We are doing “metric learning” here.

", "rawMarkdown": "We are doing “metric learning” here.", "votes": null }, { "id": "682788", "postDate": "11/27/2019 21:08:35", "content": "

\"\"\nMe assembling the pieces for this competition

", "rawMarkdown": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1182060%2Fc3265ff6010e22d1d6cac9891076e7be%2FScreenshot%20from%202019-11-27%2018-06-48.png?generation=1574888856541296&alt=media)\n`Me assembling the pieces for this competition`", "votes": null }, { "id": "683196", "postDate": "11/28/2019 07:28:47", "content": "

They better to give a third kind of prize for the one finding the correct metric😂

", "rawMarkdown": "They better to give a third kind of prize for the one finding the correct metric😂", "votes": null }, { "id": "684678", "postDate": "11/30/2019 08:42:19", "content": "

My understanding of F1 score: see https://www.youtube.com/watch?v=yIdF-17HwSk : at times 28:16 and 37:00: basically, bag of words; exact match of span indices is probably not essential

", "rawMarkdown": "My understanding of F1 score: see https://www.youtube.com/watch?v=yIdF-17HwSk : at times 28:16 and 37:00: basically, bag of words; exact match of span indices is probably not essential", "votes": null } ], "comments": [ { "id": 682287, "author_name": "lopuhin", "author_url": "", "post_date": "11/27/2019 07:27:04", "content": "

Nice summary of all recent competitions with a custom metric on Kaggle. Maybe you could also add \"waiting for metric to be fixed\" somewhere :)

\n\n

Seriously, it would be great if all competitions with a custom metric had actual metric code published (not just the inner loop like in https://www.kaggle.com/c/pku-autonomous-driving/overview/evaluation). Even better if it's possible to move to python metric implementation, this may also reduce amount of bugs or time to fix them, because participants will be able to help, and you could use existing metric implementations more.

", "votes": null, "replies": [] }, { "id": 682298, "author_name": "kashnitsky", "author_url": "", "post_date": "11/27/2019 08:06:02", "content": "

We are doing “metric learning” here.

", "votes": null, "replies": [ { "id": 683196, "author_name": "yihdarshieh", "author_url": "", "post_date": "11/28/2019 07:28:47", "content": "

They better to give a third kind of prize for the one finding the correct metric😂

", "votes": null, "replies": [] } ] }, { "id": 682788, "author_name": "dimitreoliveira", "author_url": "", "post_date": "11/27/2019 21:08:35", "content": "

\"\"\nMe assembling the pieces for this competition

", "votes": null, "replies": [] }, { "id": 684678, "author_name": "arvindpdmn", "author_url": "", "post_date": "11/30/2019 08:42:19", "content": "

My understanding of F1 score: see https://www.youtube.com/watch?v=yIdF-17HwSk : at times 28:16 and 37:00: basically, bag of words; exact match of span indices is probably not essential

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "682047": "![](https://i.imgflip.com/3hn908.jpg)", "682287": "Nice summary of all recent competitions with a custom metric on Kaggle. Maybe you could also add \"waiting for metric to be fixed\" somewhere :)\n\nSeriously, it would be great if all competitions with a custom metric had actual metric code published (not just the inner loop like in https://www.kaggle.com/c/pku-autonomous-driving/overview/evaluation). Even better if it's possible to move to python metric implementation, this may also reduce amount of bugs or time to fix them, because participants will be able to help, and you could use existing metric implementations more.", "682298": "We are doing “metric learning” here.", "682788": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1182060%2Fc3265ff6010e22d1d6cac9891076e7be%2FScreenshot%20from%202019-11-27%2018-06-48.png?generation=1574888856541296&alt=media)\n`Me assembling the pieces for this competition`", "683196": "They better to give a third kind of prize for the one finding the correct metric😂", "684678": "My understanding of F1 score: see https://www.youtube.com/watch?v=yIdF-17HwSk : at times 28:16 and 37:00: basically, bag of words; exact match of span indices is probably not essential" }, "source": "meta" }