{ "id": 237374, "title": "Potential for shake up", "url": "/competitions/hpa-single-cell-image-classification/discussion/237374", "author_name": "narsil (jobs-in-data.com)", "post_date": "2021-05-08T12:13:53.901000", "votes": 17, "comment_count": 27, "views": 0, "content": "

Well, what do you think? :)

", "messages": [ { "id": 1297897, "postDate": "2021-05-08T12:13:53.903Z", "content": "

Well, what do you think? :)

", "rawMarkdown": "Well, what do you think? :)", "votes": 17 }, { "id": 1298307, "postDate": "2021-05-08T18:04:58.560Z", "content": "

Hard to say. But one thing its going to be sure: I am gonna rest a bit next weaks. Did not have so much sleep during the last 3 month 😃

", "rawMarkdown": "Hard to say. But one thing its going to be sure: I am gonna rest a bit next weaks. Did not have so much sleep during the last 3 month 😃", "votes": 13, "replies": [ { "id": 1298334, "postDate": "2021-05-08T18:31:59.180Z", "content": "

I fully empathize with you

", "rawMarkdown": "I fully empathize with you", "votes": 2 }, { "id": 1298346, "postDate": "2021-05-08T18:48:04.417Z", "content": "

Same. Managing office work and competitions has been a pain.

", "rawMarkdown": "Same. Managing office work and competitions has been a pain.", "votes": 4 }, { "id": 1298353, "postDate": "2021-05-08T18:59:12.147Z", "content": "

Yes - like taking leave from work to work on this competition, or waking up 5 am to check your experiments, while on the holiday trip with your family. But hey - the beauty of mitotic spindle makes up for all that :)

", "rawMarkdown": "Yes - like taking leave from work to work on this competition, or waking up 5 am to check your experiments, while on the holiday trip with your family. But hey - the beauty of mitotic spindle makes up for all that :)", "votes": 7 }, { "id": 1298437, "postDate": "2021-05-08T21:13:53.143Z", "content": "

Same here! I wish we could meet after the final deadline and celebrate :) Good luck everyone!

", "rawMarkdown": "Same here! I wish we could meet after the final deadline and celebrate :) Good luck everyone!", "votes": 10 }, { "id": 1299535, "postDate": "2021-05-09T18:38:09.960Z", "content": "

while the competition may end soon, do keep us updated on your \"magic\" @thedrcat and how she develops into a future Kaggle GM!! 😄

\n

All the best everyone! hope we all are in the right side of any shakeup!!!
\nGiven 4 competitions ending soon, I am only working on this and Indoor… will be relaxed when these two end and get some sleep!!

", "rawMarkdown": "while the competition may end soon, do keep us updated on your \"magic\" @thedrcat and how she develops into a future Kaggle GM!! 😄\n\nAll the best everyone! hope we all are in the right side of any shakeup!!! \nGiven 4 competitions ending soon, I am only working on this and Indoor... will be relaxed when these two end and get some sleep!!\n", "votes": 2 } ] }, { "id": 1297939, "postDate": "2021-05-08T12:44:36.270Z", "content": "

Maybe we would call it shuffle rather than shake up 3 days later.

", "rawMarkdown": "Maybe we would call it shuffle rather than shake up 3 days later.", "votes": 5, "replies": [ { "id": 1298558, "postDate": "2021-05-09T02:41:51.063Z", "content": "

There may be a new Kaggle Competition Grandmaster at the end of this competition.😃

", "rawMarkdown": "There may be a new Kaggle Competition Grandmaster at the end of this competition.😃", "votes": 2 }, { "id": 1298646, "postDate": "2021-05-09T05:19:02.717Z", "content": "

Let's hope!

", "rawMarkdown": "Let's hope!", "votes": 3 } ] }, { "id": 1298814, "postDate": "2021-05-09T08:23:47.010Z", "content": "

I think…. Yes.

\n
\n

In the hidden test set of this challenge, we purposely chose images with high single cell variation (SCV). Therefore I believe you won’t gain significant advantages with a metric learning approach (like the winning solution last challenge)

\n

For this competition, we specifically annotated every single cell in a subset of images (mostly private/never published, and some public images) with higher SCV compared to public HPA images, to act as test set.
\n In other words, the test set will have higher SCV than the training set.

\n
\n
 **There are NO images in the private test set that were ever published.**\n
\n

Source
\nhttps://www.kaggle.com/lnhtrang/single-cell-patterns

", "rawMarkdown": "I think.... Yes. \n\n\n> In the **hidden test set** of this challenge, we purposely chose images with high single cell variation (SCV). Therefore I believe you won’t gain significant advantages with a metric learning approach (like the winning solution last challenge)\n\n> For this competition, we specifically annotated every single cell in a subset of images **(mostly private/never published**, and some public images) with higher SCV compared to public HPA images, to act as test set. \nIn other words, **the test set will have higher SCV than the training set.**\n\n```\n **There are NO images in the private test set that were ever published.**\n\n```\n**Source**\nhttps://www.kaggle.com/lnhtrang/single-cell-patterns\n", "votes": 3, "replies": [ { "id": 1298830, "postDate": "2021-05-09T08:48:41.770Z", "content": "

My intuition is very similar

", "rawMarkdown": "My intuition is very similar" }, { "id": 1298869, "postDate": "2021-05-09T09:30:12.437Z", "content": "

We have 717 teams

\n

~ 300 of them not active.
\n~ 300 of them submitting public set only ( without knowing) :)

\n

so I expect the shake up to be happening for the top ~ 117 teams

", "rawMarkdown": "We have 717 teams \n\n~ 300 of them not active.\n~ 300 of them submitting public set only ( without knowing) :)\n\nso I expect the shake up to be happening for the top ~ 117 teams ", "votes": 2 } ] }, { "id": 1298111, "postDate": "2021-05-08T15:00:19.347Z", "content": "

Given the challenges involved in doing proper cross-validation offline there could be a tendency for many teams to over -optimize for the public test set.

\n

Been a great competition regardless - good luck to all.

", "rawMarkdown": "Given the challenges involved in doing proper cross-validation offline there could be a tendency for many teams to over -optimize for the public test set. \n\nBeen a great competition regardless - good luck to all.", "votes": 3 }, { "id": 1298231, "postDate": "2021-05-08T16:58:00.970Z", "content": "

Number of data in testset is a bit low. Shake-up could happen. I hope it will not.

", "rawMarkdown": "Number of data in testset is a bit low. Shake-up could happen. I hope it will not.", "votes": 1 }, { "id": 1297978, "postDate": "2021-05-08T13:25:07.907Z", "content": "

I think there might be a slight shake up , if any.. Mitotic spindle seems to be the biggest factor that will cause a shake up , due to it being almost non existent in the training set - however , it seems to be covered well enough with external data. But you never know- I get Bengali.AI flashbacks often 😣

", "rawMarkdown": "I think there might be a slight shake up , if any.. Mitotic spindle seems to be the biggest factor that will cause a shake up , due to it being almost non existent in the training set - however , it seems to be covered well enough with external data. But you never know- I get Bengali.AI flashbacks often 😣", "votes": 1, "replies": [ { "id": 1297984, "postDate": "2021-05-08T13:28:11.790Z", "content": "

I got Microsoft malware flashback when I see shake up.

", "rawMarkdown": "I got Microsoft malware flashback when I see shake up.", "votes": 4 }, { "id": 1297990, "postDate": "2021-05-08T13:30:35.673Z", "content": "

I had no idea about that competition - just saw your profile , damn that shake up was something…

", "rawMarkdown": "I had no idea about that competition - just saw your profile , damn that shake up was something...", "votes": 1 }, { "id": 1298072, "postDate": "2021-05-08T14:33:18.920Z", "content": "

Does your model's predictions for mitotic spindle become a lot better with external data? My single cell models don't seem to show much improvement.

", "rawMarkdown": "Does your model's predictions for mitotic spindle become a lot better with external data? My single cell models don't seem to show much improvement." }, { "id": 1298152, "postDate": "2021-05-08T15:45:18.380Z", "content": "

I am not really sure to what extent the external data improves it. However , I am certain that it does improve it at least by 20%. Since the external data consists of at least a few hundred samples on mitotic spindle , you will most likely get a good improvement for that class as long as you have a good validation split.

", "rawMarkdown": "I am not really sure to what extent the external data improves it. However , I am certain that it does improve it at least by 20%. Since the external data consists of at least a few hundred samples on mitotic spindle , you will most likely get a good improvement for that class as long as you have a good validation split. ", "votes": 1 } ] }, { "id": 1302837, "postDate": "2021-05-11T18:37:35.170Z", "content": "

@narsil, I root for your team, so I wish you no shake-up. Fingers crossed 😊

", "rawMarkdown": "@narsil, I root for your team, so I wish you no shake-up. Fingers crossed 😊", "votes": 2, "replies": [ { "id": 1303164, "postDate": "2021-05-12T00:12:16.633Z", "content": "

It did happen but to a very reasonable extent :) The next stop is in Pimiento )

", "rawMarkdown": "It did happen but to a very reasonable extent :) The next stop is in Pimiento )" } ] }, { "id": 1298101, "postDate": "2021-05-08T14:54:59.667Z", "content": "

This has been an intense competition, good luck to everyone 🤞

", "rawMarkdown": "This has been an intense competition, good luck to everyone 🤞", "votes": 2, "replies": [ { "id": 1301174, "postDate": "2021-05-11T01:12:39.047Z", "content": "

great competition so far, enjoying this competition as we learn many things from great minds :D

", "rawMarkdown": "great competition so far, enjoying this competition as we learn many things from great minds :D", "votes": 1 } ] }, { "id": 1298081, "postDate": "2021-05-08T14:39:17.130Z", "content": "

This is my first contest and first shape up… hope it won't be too shaky :D No idea what to expect…

", "rawMarkdown": "This is my first contest and first shape up... hope it won't be too shaky :D No idea what to expect...", "votes": 2, "replies": [ { "id": 1298098, "postDate": "2021-05-08T14:52:46.570Z", "content": "

Love and peace!

", "rawMarkdown": "Love and peace!", "votes": 3 }, { "id": 1298102, "postDate": "2021-05-08T14:55:17.367Z", "content": "

❤️ ❤️ ❤️ ❤️ ❤️ ❤️

", "rawMarkdown": "❤️ ❤️ ❤️ ❤️ ❤️ ❤️ ", "votes": 1 } ] }, { "id": 1300657, "postDate": "2021-05-10T16:03:14.197Z", "rawMarkdown": "", "votes": -2, "isDeleted": true } ], "comments": [ { "id": 1298307, "author_name": "LucaMTB", "author_url": "", "post_date": "2021-05-08T18:04:58.560000", "content": "

Hard to say. But one thing its going to be sure: I am gonna rest a bit next weaks. Did not have so much sleep during the last 3 month 😃

", "votes": 13, "replies": [ { "id": 1298334, "author_name": "narsil (jobs-in-data.com)", "author_url": "", "post_date": "2021-05-08T18:31:59.180000", "content": "

I fully empathize with you

", "votes": 2, "replies": [] }, { "id": 1298346, "author_name": "Satwik", "author_url": "", "post_date": "2021-05-08T18:48:04.417000", "content": "

Same. Managing office work and competitions has been a pain.

", "votes": 4, "replies": [] }, { "id": 1298353, "author_name": "narsil (jobs-in-data.com)", "author_url": "", "post_date": "2021-05-08T18:59:12.147000", "content": "

Yes - like taking leave from work to work on this competition, or waking up 5 am to check your experiments, while on the holiday trip with your family. But hey - the beauty of mitotic spindle makes up for all that :)

", "votes": 7, "replies": [] }, { "id": 1298437, "author_name": "Darek Kłeczek", "author_url": "", "post_date": "2021-05-08T21:13:53.143000", "content": "

Same here! I wish we could meet after the final deadline and celebrate :) Good luck everyone!

", "votes": 10, "replies": [] }, { "id": 1299535, "author_name": "Kamal Das", "author_url": "", "post_date": "2021-05-09T18:38:09.960000", "content": "

while the competition may end soon, do keep us updated on your \"magic\" @thedrcat and how she develops into a future Kaggle GM!! 😄

\n

All the best everyone! hope we all are in the right side of any shakeup!!!
\nGiven 4 competitions ending soon, I am only working on this and Indoor… will be relaxed when these two end and get some sleep!!

", "votes": 2, "replies": [] } ] }, { "id": 1297939, "author_name": "sheep", "author_url": "", "post_date": "2021-05-08T12:44:36.270000", "content": "

Maybe we would call it shuffle rather than shake up 3 days later.

", "votes": 5, "replies": [ { "id": 1298558, "author_name": "Alien", "author_url": "", "post_date": "2021-05-09T02:41:51.063000", "content": "

There may be a new Kaggle Competition Grandmaster at the end of this competition.😃

", "votes": 2, "replies": [] }, { "id": 1298646, "author_name": "sin", "author_url": "", "post_date": "2021-05-09T05:19:02.717000", "content": "

Let's hope!

", "votes": 3, "replies": [] } ] }, { "id": 1298814, "author_name": "Faisal Alsrheed", "author_url": "", "post_date": "2021-05-09T08:23:47.010000", "content": "

I think…. Yes.

\n
\n

In the hidden test set of this challenge, we purposely chose images with high single cell variation (SCV). Therefore I believe you won’t gain significant advantages with a metric learning approach (like the winning solution last challenge)

\n

For this competition, we specifically annotated every single cell in a subset of images (mostly private/never published, and some public images) with higher SCV compared to public HPA images, to act as test set.
\n In other words, the test set will have higher SCV than the training set.

\n
\n
 **There are NO images in the private test set that were ever published.**\n
\n

Source
\nhttps://www.kaggle.com/lnhtrang/single-cell-patterns

", "votes": 3, "replies": [ { "id": 1298830, "author_name": "narsil (jobs-in-data.com)", "author_url": "", "post_date": "2021-05-09T08:48:41.770000", "content": "

My intuition is very similar

", "votes": 0, "replies": [] }, { "id": 1298869, "author_name": "Faisal Alsrheed", "author_url": "", "post_date": "2021-05-09T09:30:12.437000", "content": "

We have 717 teams

\n

~ 300 of them not active.
\n~ 300 of them submitting public set only ( without knowing) :)

\n

so I expect the shake up to be happening for the top ~ 117 teams

", "votes": 2, "replies": [] } ] }, { "id": 1298111, "author_name": "Andrew Tratz", "author_url": "", "post_date": "2021-05-08T15:00:19.347000", "content": "

Given the challenges involved in doing proper cross-validation offline there could be a tendency for many teams to over -optimize for the public test set.

\n

Been a great competition regardless - good luck to all.

", "votes": 3, "replies": [] }, { "id": 1298231, "author_name": "MPWARE", "author_url": "", "post_date": "2021-05-08T16:58:00.970000", "content": "

Number of data in testset is a bit low. Shake-up could happen. I hope it will not.

", "votes": 1, "replies": [] }, { "id": 1297978, "author_name": "Satwik", "author_url": "", "post_date": "2021-05-08T13:25:07.907000", "content": "

I think there might be a slight shake up , if any.. Mitotic spindle seems to be the biggest factor that will cause a shake up , due to it being almost non existent in the training set - however , it seems to be covered well enough with external data. But you never know- I get Bengali.AI flashbacks often 😣

", "votes": 1, "replies": [ { "id": 1297984, "author_name": "sheep", "author_url": "", "post_date": "2021-05-08T13:28:11.790000", "content": "

I got Microsoft malware flashback when I see shake up.

", "votes": 4, "replies": [] }, { "id": 1297990, "author_name": "Satwik", "author_url": "", "post_date": "2021-05-08T13:30:35.673000", "content": "

I had no idea about that competition - just saw your profile , damn that shake up was something…

", "votes": 1, "replies": [] }, { "id": 1298072, "author_name": "novice03", "author_url": "", "post_date": "2021-05-08T14:33:18.920000", "content": "

Does your model's predictions for mitotic spindle become a lot better with external data? My single cell models don't seem to show much improvement.

", "votes": 0, "replies": [] }, { "id": 1298152, "author_name": "Satwik", "author_url": "", "post_date": "2021-05-08T15:45:18.380000", "content": "

I am not really sure to what extent the external data improves it. However , I am certain that it does improve it at least by 20%. Since the external data consists of at least a few hundred samples on mitotic spindle , you will most likely get a good improvement for that class as long as you have a good validation split.

", "votes": 1, "replies": [] } ] }, { "id": 1302837, "author_name": "Alvor", "author_url": "", "post_date": "2021-05-11T18:37:35.170000", "content": "

@narsil, I root for your team, so I wish you no shake-up. Fingers crossed 😊

", "votes": 2, "replies": [ { "id": 1303164, "author_name": "narsil (jobs-in-data.com)", "author_url": "", "post_date": "2021-05-12T00:12:16.633000", "content": "

It did happen but to a very reasonable extent :) The next stop is in Pimiento )

", "votes": 0, "replies": [] } ] }, { "id": 1298101, "author_name": "novice03", "author_url": "", "post_date": "2021-05-08T14:54:59.667000", "content": "

This has been an intense competition, good luck to everyone 🤞

", "votes": 2, "replies": [ { "id": 1301174, "author_name": "Acmd Gani", "author_url": "", "post_date": "2021-05-11T01:12:39.047000", "content": "

great competition so far, enjoying this competition as we learn many things from great minds :D

", "votes": 1, "replies": [] } ] }, { "id": 1298081, "author_name": "CroDoc", "author_url": "", "post_date": "2021-05-08T14:39:17.130000", "content": "

This is my first contest and first shape up… hope it won't be too shaky :D No idea what to expect…

", "votes": 2, "replies": [ { "id": 1298098, "author_name": "sheep", "author_url": "", "post_date": "2021-05-08T14:52:46.570000", "content": "

Love and peace!

", "votes": 3, "replies": [] }, { "id": 1298102, "author_name": "CroDoc", "author_url": "", "post_date": "2021-05-08T14:55:17.367000", "content": "

❤️ ❤️ ❤️ ❤️ ❤️ ❤️

", "votes": 1, "replies": [] } ] }, { "id": 1300657, "author_name": "", "author_url": "", "post_date": "2021-05-10T16:03:14.197000", "content": "", "votes": -2, "replies": [] } ], "raw_markdown_by_id": { "1297897": "Well, what do you think? :)", "1298307": "Hard to say. But one thing its going to be sure: I am gonna rest a bit next weaks. Did not have so much sleep during the last 3 month 😃", "1297939": "Maybe we would call it shuffle rather than shake up 3 days later.", "1298814": "I think.... Yes. \n\n\n> In the **hidden test set** of this challenge, we purposely chose images with high single cell variation (SCV). Therefore I believe you won’t gain significant advantages with a metric learning approach (like the winning solution last challenge)\n\n> For this competition, we specifically annotated every single cell in a subset of images **(mostly private/never published**, and some public images) with higher SCV compared to public HPA images, to act as test set. \nIn other words, **the test set will have higher SCV than the training set.**\n\n```\n **There are NO images in the private test set that were ever published.**\n\n```\n**Source**\nhttps://www.kaggle.com/lnhtrang/single-cell-patterns\n", "1298111": "Given the challenges involved in doing proper cross-validation offline there could be a tendency for many teams to over -optimize for the public test set. \n\nBeen a great competition regardless - good luck to all.", "1298231": "Number of data in testset is a bit low. Shake-up could happen. I hope it will not.", "1297978": "I think there might be a slight shake up , if any.. Mitotic spindle seems to be the biggest factor that will cause a shake up , due to it being almost non existent in the training set - however , it seems to be covered well enough with external data. But you never know- I get Bengali.AI flashbacks often 😣", "1302837": "@narsil, I root for your team, so I wish you no shake-up. Fingers crossed 😊", "1298101": "This has been an intense competition, good luck to everyone 🤞", "1298081": "This is my first contest and first shape up... hope it won't be too shaky :D No idea what to expect...", "1300657": "" } }