![\"screenshot\"](\"https://i.ibb.co/KyWfp0y/Screen-Shot-2022-03-12-at-10-55-59-AM.png\")
With the new clarification that pretrained MNIST models are not allowed shouldn't the lb be reset ? Because this is a new rule as previously it was stated that using publicly available MNIST weights will be Allowed. I am sure the top solutions currently use pretrained MNIST weights
", "messages": [ { "id": "1720208", "postDate": "03/12/2022 15:01:42", "content": "With the new clarification that pretrained MNIST models are not allowed shouldn't the lb be reset ? Because this is a new rule as previously it was stated that using publicly available MNIST weights will be Allowed. I am sure the top solutions currently use pretrained MNIST weights
", "rawMarkdown": "With the new clarification that pretrained MNIST models are not allowed shouldn't the lb be reset ? Because this is a new rule as previously it was stated that using publicly available MNIST weights will be Allowed. I am sure the top solutions currently use pretrained MNIST weights", "votes": null }, { "id": "1720252", "postDate": "03/12/2022 15:55:35", "content": "While I am abstaining from participating in the competition further, I will comment and say that all my submissions used pre-trained MNIST models.
\nI changed my team name in this competition to have include 🚨🚨🚨 as a prefix and suffix. Hopefully, this will help others identify submissions using MNIST models. This obviously only helps if others use this technique though.
\nSee below:
\n
Thank you @dschettler8845 for the great idea.
", "rawMarkdown": "Thank you @dschettler8845 for the great idea.", "votes": null }, { "id": "1720285", "postDate": "03/12/2022 16:40:45", "content": "@dschettler8845 We are really sorry to hear that you are not participating any further. This competition, with its constrained setting, would have benefitted from Kagglers like you. We also appreciate your idea of putting prefix and suffix around MNIST solutions. If you do not mind, we would like to announce this as a general rule for those interested in testing general solutions (beyond the GPU track rules).
\nOn a side note, we have added the constraints on the problem only to maintain the essence. Through this challenge, we are initiating a whole new direction of research for some realistic scenarios. We agree the constraints should be have been defined earlier, and we will take this into account in our next competition (a fullscale challenge on Kaggle or another platform as a follow up on this) that is already in planning.
", "rawMarkdown": "dschettler8845 We are really sorry to hear that you are not participating any further. This competition, with its constrained setting, would have benefitted from Kagglers like you. We also appreciate your idea of putting prefix and suffix around MNIST solutions. If you do not mind, we would like to announce this as a general rule for those interested in testing general solutions (beyond the GPU track rules).\n\nOn a side note, we have added the constraints on the problem only to maintain the essence. Through this challenge, we are initiating a whole new direction of research for some realistic scenarios. We agree the constraints should be have been defined earlier, and we will take this into account in our next competition (a fullscale challenge on Kaggle or another platform as a follow up on this) that is already in planning.", "votes": null }, { "id": "1720293", "postDate": "03/12/2022 16:54:11", "content": "Thanks for the comment @dkgupta90 !
\nYou can absolutely use that idea. I think if enough people do it, it will mitigate some of the LB confusion. Consequently, you could ask those that are submitting W/O using pretrained MNIST models to add something to their team name (then you could simply see how your score compares to others who have the 🔬 symbol or something)? Or both!
\nI think this is a really cool competition and I really wish, as you had said, it had been defined a bit clearer earlier on. Regardless, I appreciate your's and the team's dedication to generating interesting new challenges and I will keep an eye out for future competitions that encourage this type of lateral thinking.
\nPerhaps if I had less on my plate right now I would be more open to trying again. Regardless, I really hope the competition goes well for the hosts and everyone else! I'll be keen to read about the solutions when this wraps!
\nAll the best!
", "rawMarkdown": "Thanks for the comment @dkgupta90 !\n\nYou can absolutely use that idea. I think if enough people do it, it will mitigate some of the LB confusion. Consequently, you could ask those that are submitting W/O using pretrained MNIST models to add something to their team name (then you could simply see how your score compares to others who have the 🔬 symbol or something)? Or both!\n\nI think this is a really cool competition and I really wish, as you had said, it had been defined a bit clearer earlier on. Regardless, I appreciate your's and the team's dedication to generating interesting new challenges and I will keep an eye out for future competitions that encourage this type of lateral thinking.\n\nPerhaps if I had less on my plate right now I would be more open to trying again. Regardless, I really hope the competition goes well for the hosts and everyone else! I'll be keen to read about the solutions when this wraps!\n\nAll the best!", "votes": null }, { "id": "1722642", "postDate": "03/14/2022 18:38:24", "content": "Everything was getting messy so we decided to make the data more complex and alleviate the rules. If Kagglers still remove the background (which we hope not), we can't do much about it. Lets see.
", "rawMarkdown": "Everything was getting messy so we decided to make the data more complex and alleviate the rules. If Kagglers still remove the background (which we hope not), we can't do much about it. Lets see.", "votes": null } ], "comments": [ { "id": 1720252, "author_name": "dschettler8845", "author_url": "", "post_date": "03/12/2022 15:55:35", "content": "While I am abstaining from participating in the competition further, I will comment and say that all my submissions used pre-trained MNIST models.
\nI changed my team name in this competition to have include 🚨🚨🚨 as a prefix and suffix. Hopefully, this will help others identify submissions using MNIST models. This obviously only helps if others use this technique though.
\nSee below:
\n
Thank you @dschettler8845 for the great idea.
", "votes": null, "replies": [] }, { "id": 1720285, "author_name": "dkgupta90", "author_url": "", "post_date": "03/12/2022 16:40:45", "content": "@dschettler8845 We are really sorry to hear that you are not participating any further. This competition, with its constrained setting, would have benefitted from Kagglers like you. We also appreciate your idea of putting prefix and suffix around MNIST solutions. If you do not mind, we would like to announce this as a general rule for those interested in testing general solutions (beyond the GPU track rules).
\nOn a side note, we have added the constraints on the problem only to maintain the essence. Through this challenge, we are initiating a whole new direction of research for some realistic scenarios. We agree the constraints should be have been defined earlier, and we will take this into account in our next competition (a fullscale challenge on Kaggle or another platform as a follow up on this) that is already in planning.
", "votes": null, "replies": [] }, { "id": 1720293, "author_name": "dschettler8845", "author_url": "", "post_date": "03/12/2022 16:54:11", "content": "Thanks for the comment @dkgupta90 !
\nYou can absolutely use that idea. I think if enough people do it, it will mitigate some of the LB confusion. Consequently, you could ask those that are submitting W/O using pretrained MNIST models to add something to their team name (then you could simply see how your score compares to others who have the 🔬 symbol or something)? Or both!
\nI think this is a really cool competition and I really wish, as you had said, it had been defined a bit clearer earlier on. Regardless, I appreciate your's and the team's dedication to generating interesting new challenges and I will keep an eye out for future competitions that encourage this type of lateral thinking.
\nPerhaps if I had less on my plate right now I would be more open to trying again. Regardless, I really hope the competition goes well for the hosts and everyone else! I'll be keen to read about the solutions when this wraps!
\nAll the best!
", "votes": null, "replies": [] }, { "id": 1722642, "author_name": "dkgupta90", "author_url": "", "post_date": "03/14/2022 18:38:24", "content": "Everything was getting messy so we decided to make the data more complex and alleviate the rules. If Kagglers still remove the background (which we hope not), we can't do much about it. Lets see.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1720208": "With the new clarification that pretrained MNIST models are not allowed shouldn't the lb be reset ? Because this is a new rule as previously it was stated that using publicly available MNIST weights will be Allowed. I am sure the top solutions currently use pretrained MNIST weights", "1720252": "While I am abstaining from participating in the competition further, I will comment and say that all my submissions used pre-trained MNIST models. \n\nI changed my team name in this competition to have include 🚨🚨🚨 as a prefix and suffix. Hopefully, this will help others identify submissions using MNIST models. This obviously only helps if others use this technique though.\n\nSee below:\n\n", "1720264": "Thank you @dschettler8845 for the great idea.", "1720285": "dschettler8845 We are really sorry to hear that you are not participating any further. This competition, with its constrained setting, would have benefitted from Kagglers like you. We also appreciate your idea of putting prefix and suffix around MNIST solutions. If you do not mind, we would like to announce this as a general rule for those interested in testing general solutions (beyond the GPU track rules).\n\nOn a side note, we have added the constraints on the problem only to maintain the essence. Through this challenge, we are initiating a whole new direction of research for some realistic scenarios. We agree the constraints should be have been defined earlier, and we will take this into account in our next competition (a fullscale challenge on Kaggle or another platform as a follow up on this) that is already in planning.", "1720293": "Thanks for the comment @dkgupta90 !\n\nYou can absolutely use that idea. I think if enough people do it, it will mitigate some of the LB confusion. Consequently, you could ask those that are submitting W/O using pretrained MNIST models to add something to their team name (then you could simply see how your score compares to others who have the 🔬 symbol or something)? Or both!\n\nI think this is a really cool competition and I really wish, as you had said, it had been defined a bit clearer earlier on. Regardless, I appreciate your's and the team's dedication to generating interesting new challenges and I will keep an eye out for future competitions that encourage this type of lateral thinking.\n\nPerhaps if I had less on my plate right now I would be more open to trying again. Regardless, I really hope the competition goes well for the hosts and everyone else! I'll be keen to read about the solutions when this wraps!\n\nAll the best!", "1722642": "Everything was getting messy so we decided to make the data more complex and alleviate the rules. If Kagglers still remove the background (which we hope not), we can't do much about it. Lets see." }, "source": "meta" }