Thanks to Kaggle, California Healthcare Foundation and EyePACS for sponsoring an interesting and worthwhile competition. Barring very unusual re-shuffling of leaderboard scores, it looks like the machine learning scholar from the U.K. will win, using some sort of custom algorithm that pushes forward the state-of-the-art in deep learning. I look forward to learning more about it. It also appears that the winning algorithms have achieved the goal of generating ophthalmologist-level diagnoses, as in the computer-to-human disagreement appears similar to what I would expect of human-to-human disagreement. Overall this competition seems to have been a great success.
", "messages": [ { "id": "87225", "postDate": "07/27/2015 22:26:27", "content": "Thanks to Kaggle, California Healthcare Foundation and EyePACS for sponsoring an interesting and worthwhile competition. Barring very unusual re-shuffling of leaderboard scores, it looks like the machine learning scholar from the U.K. will win, using some sort of custom algorithm that pushes forward the state-of-the-art in deep learning. I look forward to learning more about it. It also appears that the winning algorithms have achieved the goal of generating ophthalmologist-level diagnoses, as in the computer-to-human disagreement appears similar to what I would expect of human-to-human disagreement. Overall this competition seems to have been a great success.
", "rawMarkdown": "Thanks to Kaggle, California Healthcare Foundation and EyePACS for sponsoring an interesting and worthwhile competition. Barring very unusual re-shuffling of leaderboard scores, it looks like the machine learning scholar from the U.K. will win, using some sort of custom algorithm that pushes forward the state-of-the-art in deep learning. I look forward to learning more about it. It also appears that the winning algorithms have achieved the goal of generating ophthalmologist-level diagnoses, as in the computer-to-human disagreement appears similar to what I would expect of human-to-human disagreement. Overall this competition seems to have been a great success.", "votes": null }, { "id": "87236", "postDate": "07/27/2015 23:53:32", "content": "Yep, thanks everyone! That was exciting time!
", "rawMarkdown": "Yep, thanks everyone! That was exciting time!", "votes": null }, { "id": "87241", "postDate": "07/28/2015 00:06:14", "content": "I enjoyed this competition a lot, it was fun to play with the latest in deep learning. I hope to learn some more from the winning solutions!
", "rawMarkdown": "I enjoyed this competition a lot, it was fun to play with the latest in deep learning. I hope to learn some more from the winning solutions!", "votes": null }, { "id": "87243", "postDate": "07/28/2015 00:17:29", "content": "I badly suffered from "fractional max-pooling" envy during the competition.\nI'm very curious about the winning solution.
\n\nThanks to Daniel Hammack.. It was a blast to work together!
\n\nAs for Ben Graham. \nCongrats. You are my favorite convnet/deep learning author. So many good ideas and always very readable and understandable.
\n\nI run your code and although I could not make it score well I loved reading how you solved certain problems. Great ideas and angles that I never have seen before in other convnet implementations.
", "rawMarkdown": "I badly suffered from \"fractional max-pooling\" envy during the competition.\r\nI'm very curious about the winning solution.\r\n\r\nThanks to Daniel Hammack.. It was a blast to work together!\r\n\r\nAs for Ben Graham. \r\nCongrats. You are my favorite convnet/deep learning author. So many good ideas and always very readable and understandable.\r\n\r\nI run your code and although I could not make it score well I loved reading how you solved certain problems. Great ideas and angles that I never have seen before in other convnet implementations.", "votes": null }, { "id": "87244", "postDate": "07/28/2015 00:24:44", "content": "Same, I learned a lot!
\n\nCongrats to the winners! :-)
", "rawMarkdown": "Same, I learned a lot!\r\n\r\nCongrats to the winners! :-)", "votes": null }, { "id": "87247", "postDate": "07/28/2015 00:29:09", "content": "Congratulations to the winners, and thanks to the organizers for a very interesting and challenging competition.
\n\nBTW, I notice that my submissions page only shows the public scores of my submissions. Now that the contest is over, will that page eventually be modified to show the private scores as well, as was done in previous competitions?
", "rawMarkdown": "Congratulations to the winners, and thanks to the organizers for a very interesting and challenging competition. \r\n\r\nBTW, I notice that my submissions page only shows the public scores of my submissions. Now that the contest is over, will that page eventually be modified to show the private scores as well, as was done in previous competitions?", "votes": null }, { "id": "87248", "postDate": "07/28/2015 00:38:17", "content": "Thanks for organizing this exciting competition. It's really enjoyable.\nAnd congrats to the winners! I
", "rawMarkdown": "Thanks for organizing this exciting competition. It's really enjoyable.\r\nAnd congrats to the winners! I", "votes": null } ], "comments": [ { "id": 87236, "author_name": "lupoglaz", "author_url": "", "post_date": "07/27/2015 23:53:32", "content": "Yep, thanks everyone! That was exciting time!
", "votes": null, "replies": [] }, { "id": 87241, "author_name": "dhammack", "author_url": "", "post_date": "07/28/2015 00:06:14", "content": "I enjoyed this competition a lot, it was fun to play with the latest in deep learning. I hope to learn some more from the winning solutions!
", "votes": null, "replies": [] }, { "id": 87243, "author_name": "juliandewit", "author_url": "", "post_date": "07/28/2015 00:17:29", "content": "I badly suffered from "fractional max-pooling" envy during the competition.\nI'm very curious about the winning solution.
\n\nThanks to Daniel Hammack.. It was a blast to work together!
\n\nAs for Ben Graham. \nCongrats. You are my favorite convnet/deep learning author. So many good ideas and always very readable and understandable.
\n\nI run your code and although I could not make it score well I loved reading how you solved certain problems. Great ideas and angles that I never have seen before in other convnet implementations.
", "votes": null, "replies": [] }, { "id": 87244, "author_name": "jeffreydf", "author_url": "", "post_date": "07/28/2015 00:24:44", "content": "Same, I learned a lot!
\n\nCongrats to the winners! :-)
", "votes": null, "replies": [] }, { "id": 87247, "author_name": "dslate", "author_url": "", "post_date": "07/28/2015 00:29:09", "content": "Congratulations to the winners, and thanks to the organizers for a very interesting and challenging competition.
\n\nBTW, I notice that my submissions page only shows the public scores of my submissions. Now that the contest is over, will that page eventually be modified to show the private scores as well, as was done in previous competitions?
", "votes": null, "replies": [] }, { "id": 87248, "author_name": "rbnwkt", "author_url": "", "post_date": "07/28/2015 00:38:17", "content": "Thanks for organizing this exciting competition. It's really enjoyable.\nAnd congrats to the winners! I
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "87225": "Thanks to Kaggle, California Healthcare Foundation and EyePACS for sponsoring an interesting and worthwhile competition. Barring very unusual re-shuffling of leaderboard scores, it looks like the machine learning scholar from the U.K. will win, using some sort of custom algorithm that pushes forward the state-of-the-art in deep learning. I look forward to learning more about it. It also appears that the winning algorithms have achieved the goal of generating ophthalmologist-level diagnoses, as in the computer-to-human disagreement appears similar to what I would expect of human-to-human disagreement. Overall this competition seems to have been a great success.", "87236": "Yep, thanks everyone! That was exciting time!", "87241": "I enjoyed this competition a lot, it was fun to play with the latest in deep learning. I hope to learn some more from the winning solutions!", "87243": "I badly suffered from \"fractional max-pooling\" envy during the competition.\r\nI'm very curious about the winning solution.\r\n\r\nThanks to Daniel Hammack.. It was a blast to work together!\r\n\r\nAs for Ben Graham. \r\nCongrats. You are my favorite convnet/deep learning author. So many good ideas and always very readable and understandable.\r\n\r\nI run your code and although I could not make it score well I loved reading how you solved certain problems. Great ideas and angles that I never have seen before in other convnet implementations.", "87244": "Same, I learned a lot!\r\n\r\nCongrats to the winners! :-)", "87247": "Congratulations to the winners, and thanks to the organizers for a very interesting and challenging competition. \r\n\r\nBTW, I notice that my submissions page only shows the public scores of my submissions. Now that the contest is over, will that page eventually be modified to show the private scores as well, as was done in previous competitions?", "87248": "Thanks for organizing this exciting competition. It's really enjoyable.\r\nAnd congrats to the winners! I" }, "source": "meta" }