{ "id": 197805, "title": "COLA: Contrastive learning of general purpose audio representations", "url": "/competitions/rfcx-species-audio-detection/discussion/197805", "author_name": "", "post_date": "2020-11-18T06:18:18.150480100Z", "votes": 38, "comment_count": 6, "views": 0, "content": "

Seems we have relatively small number of samples in this competition, and creating good embedding with pre-train task may be helpful.

\n

Here's a novel self-supervised learning method for general purpose audio representation learning.

\n

Paper: https://arxiv.org/abs/2010.10915
\nBlog post: https://towardsdatascience.com/paper-summary-implementation-contrastive-learning-of-general-purpose-audio-representations-f4e3cc06fcf7
\nGitHub(Official, Tensorflow): https://github.com/google-research/google-research/tree/master/cola
\nGitHub(PyTorch): https://github.com/CVxTz/COLA_pytorch

", "messages": [ { "id": "1082691", "postDate": "11/18/2020 06:18:18", "content": "

Seems we have relatively small number of samples in this competition, and creating good embedding with pre-train task may be helpful.

\n

Here's a novel self-supervised learning method for general purpose audio representation learning.

\n

Paper: https://arxiv.org/abs/2010.10915
\nBlog post: https://towardsdatascience.com/paper-summary-implementation-contrastive-learning-of-general-purpose-audio-representations-f4e3cc06fcf7
\nGitHub(Official, Tensorflow): https://github.com/google-research/google-research/tree/master/cola
\nGitHub(PyTorch): https://github.com/CVxTz/COLA_pytorch

", "rawMarkdown": "Seems we have relatively small number of samples in this competition, and creating good embedding with pre-train task may be helpful.\n\nHere's a novel self-supervised learning method for general purpose audio representation learning.\n\nPaper: https://arxiv.org/abs/2010.10915\nBlog post: https://towardsdatascience.com/paper-summary-implementation-contrastive-learning-of-general-purpose-audio-representations-f4e3cc06fcf7\nGitHub(Official, Tensorflow): https://github.com/google-research/google-research/tree/master/cola\nGitHub(PyTorch): https://github.com/CVxTz/COLA_pytorch", "votes": null }, { "id": "1082719", "postDate": "11/18/2020 07:00:03", "content": "

Happy to see you here after everything you've shared on bird competition :)
\nThank you always for the generous sharing !

", "rawMarkdown": "Happy to see you here after everything you've shared on bird competition :) \nThank you always for the generous sharing !", "votes": null }, { "id": "1082846", "postDate": "11/18/2020 09:45:59", "content": "

Not sure I'll work hard on this competition (I need to focus on my master's thesis…) but I'm quite interested in another competition of audio data (which seems similar to bird's one)

", "rawMarkdown": "Not sure I'll work hard on this competition (I need to focus on my master's thesis...) but I'm quite interested in another competition of audio data (which seems similar to bird's one)", "votes": null }, { "id": "1083077", "postDate": "11/18/2020 15:12:59", "content": "

I also have my interest in audio, and this one seems unique with the imposter explicitly provided. Good luck with the master's thesis !! I'm sure you will do well 👍

", "rawMarkdown": "I also have my interest in audio, and this one seems unique with the imposter explicitly provided. Good luck with the master's thesis !! I'm sure you will do well 👍", "votes": null }, { "id": "1083362", "postDate": "11/18/2020 22:31:54", "content": "

Hi,
\nthere is also this triplet loss paper for unsupervised feature learning.

", "rawMarkdown": "Hi,\nthere is also this [triplet loss paper](https://arxiv.org/pdf/1711.02209.pdf) for unsupervised feature learning.", "votes": null }, { "id": "1119648", "postDate": "12/20/2020 08:54:03", "content": "

@hidehisaarai1213 hi, may I ask if there is also one audio competition in the mean time? What is it? Thanks!

", "rawMarkdown": "hidehisaarai1213 hi, may I ask if there is also one audio competition in the mean time? What is it? Thanks!", "votes": null }, { "id": "1169189", "postDate": "01/25/2021 11:26:32", "content": "

If you write an ai to write your master thesis -> meanwhile you can do this competition ;) :P

", "rawMarkdown": "If you write an ai to write your master thesis -> meanwhile you can do this competition ;) :P", "votes": null } ], "comments": [ { "id": 1082719, "author_name": "nyleve", "author_url": "", "post_date": "11/18/2020 07:00:03", "content": "

Happy to see you here after everything you've shared on bird competition :)
\nThank you always for the generous sharing !

", "votes": null, "replies": [ { "id": 1082846, "author_name": "hidehisaarai1213", "author_url": "", "post_date": "11/18/2020 09:45:59", "content": "

Not sure I'll work hard on this competition (I need to focus on my master's thesis…) but I'm quite interested in another competition of audio data (which seems similar to bird's one)

", "votes": null, "replies": [] }, { "id": 1083077, "author_name": "nyleve", "author_url": "", "post_date": "11/18/2020 15:12:59", "content": "

I also have my interest in audio, and this one seems unique with the imposter explicitly provided. Good luck with the master's thesis !! I'm sure you will do well 👍

", "votes": null, "replies": [] }, { "id": 1169189, "author_name": "killimi", "author_url": "", "post_date": "01/25/2021 11:26:32", "content": "

If you write an ai to write your master thesis -> meanwhile you can do this competition ;) :P

", "votes": null, "replies": [] } ] }, { "id": 1083362, "author_name": "kyuhwa", "author_url": "", "post_date": "11/18/2020 22:31:54", "content": "

Hi,
\nthere is also this triplet loss paper for unsupervised feature learning.

", "votes": null, "replies": [] }, { "id": 1119648, "author_name": "wubinbai", "author_url": "", "post_date": "12/20/2020 08:54:03", "content": "

@hidehisaarai1213 hi, may I ask if there is also one audio competition in the mean time? What is it? Thanks!

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1082691": "Seems we have relatively small number of samples in this competition, and creating good embedding with pre-train task may be helpful.\n\nHere's a novel self-supervised learning method for general purpose audio representation learning.\n\nPaper: https://arxiv.org/abs/2010.10915\nBlog post: https://towardsdatascience.com/paper-summary-implementation-contrastive-learning-of-general-purpose-audio-representations-f4e3cc06fcf7\nGitHub(Official, Tensorflow): https://github.com/google-research/google-research/tree/master/cola\nGitHub(PyTorch): https://github.com/CVxTz/COLA_pytorch", "1082719": "Happy to see you here after everything you've shared on bird competition :) \nThank you always for the generous sharing !", "1082846": "Not sure I'll work hard on this competition (I need to focus on my master's thesis...) but I'm quite interested in another competition of audio data (which seems similar to bird's one)", "1083077": "I also have my interest in audio, and this one seems unique with the imposter explicitly provided. Good luck with the master's thesis !! I'm sure you will do well 👍", "1083362": "Hi,\nthere is also this [triplet loss paper](https://arxiv.org/pdf/1711.02209.pdf) for unsupervised feature learning.", "1119648": "hidehisaarai1213 hi, may I ask if there is also one audio competition in the mean time? What is it? Thanks!", "1169189": "If you write an ai to write your master thesis -> meanwhile you can do this competition ;) :P" }, "source": "meta" }