Do the rules prevent players from solving the problem offline and just uploading the results? How?
\nWhat if they train a NN on the results and upload the weights instead? (In practice, this makes no difference, since NNs tend to memorize the training data)
", "messages": [ { "id": "1012437", "postDate": "09/16/2020 05:12:44", "content": "Do the rules prevent players from solving the problem offline and just uploading the results? How?
\nWhat if they train a NN on the results and upload the weights instead? (In practice, this makes no difference, since NNs tend to memorize the training data)
", "rawMarkdown": "Do the rules prevent players from solving the problem offline and just uploading the results? How?\n\nWhat if they train a NN on the results and upload the weights instead? (In practice, this makes no difference, since NNs tend to memorize the training data)", "votes": null }, { "id": "1012800", "postDate": "09/16/2020 09:45:43", "content": "it is unclear from the challenge 'overview' tab, but as it is a code competition, I guess your code will be re-run against unseen grids to compute the final private leaderboard.
\n@inversion or someone from kaggle team : can you clarify this point?
", "rawMarkdown": "it is unclear from the challenge 'overview' tab, but as it is a code competition, I guess your code will be re-run against unseen grids to compute the final private leaderboard.\n\n@inversion or someone from kaggle team : can you clarify this point?", "votes": null }, { "id": "1013228", "postDate": "09/16/2020 15:17:16", "content": "@olegtrott Great question, and it can be confusing with the different formats we have.
\nThis is a Code competition in the sense that a Notebook must be submitted. This is the only way that we can trigger the Game of Life solver on your solution.
\nBut your notebook can be anything from: 1) full training and inference, to 2) just inference from pretrained weights, and even just 3) a Notebook that reads a csv submission from a private dataset and submits that.
\nIn other words, you can do everything offline, even the predictions, and then have the Notebook just read/submit that prediction file.
\nHave fun and good luck!
", "rawMarkdown": "olegtrott Great question, and it can be confusing with the different formats we have.\n\nThis is a Code competition in the sense that a Notebook must be submitted. This is the only way that we can trigger the Game of Life solver on your solution.\n\nBut your notebook can be anything from: 1) full training and inference, to 2) just inference from pretrained weights, and even just 3) a Notebook that reads a csv submission from a private dataset and submits that.\n\nIn other words, you can do everything offline, even the predictions, and then have the Notebook just read/submit that prediction file.\n\nHave fun and good luck!", "votes": null }, { "id": "1013798", "postDate": "09/17/2020 01:49:48", "content": "I believe in this competition the leaderboard data is the same as the public test dataset.
\nIt would be nice if somebody on the Kaggle team could confirm this, and if this applies to both the public and the private leaderboards.
", "rawMarkdown": "I believe in this competition the leaderboard data is the same as the public test dataset.\n\nIt would be nice if somebody on the Kaggle team could confirm this, and if this applies to both the public and the private leaderboards.", "votes": null }, { "id": "1028534", "postDate": "09/27/2020 00:57:50", "content": "Follow-up Q: where can we find resource and time constraints for the code execution? (Nevermind, found it: https://www.kaggle.com/c/conways-reverse-game-of-life-2020/overview/code-requirements )
", "rawMarkdown": "Follow-up Q: where can we find resource and time constraints for the code execution? (Nevermind, found it: https://www.kaggle.com/c/conways-reverse-game-of-life-2020/overview/code-requirements )", "votes": null }, { "id": "1045987", "postDate": "10/11/2020 08:12:34", "content": "I am having some issues getting that to work. I have uploaded my submission.csv file and tried triggering the kernel for registering it as an output but keep getting an error. My notebook has
\ntbl <- read.csv(file = file.path(\"..\", \"input\", \"rivanna\", \"data\", \"submission.csv\"))\nwrite.csv(x = tbl, file = \"submission.csv\")\nwrite.csv(x = tbl, file = file.path(\"..\", \"submission.csv\"))\nThe submission.csv shows up after committing the notebook but keep getting the
Submission Scoring Error\nEvaluation Exception: No submission file found. Expected submission.csv\nYes, and it seems as if the leaderboard data that is being displayed now is only for half of the grids that are in the test set, is this correct?
", "rawMarkdown": "Yes, and it seems as if the leaderboard data that is being displayed now is only for half of the grids that are in the test set, is this correct?", "votes": null } ], "comments": [ { "id": 1012800, "author_name": "ebouteillon", "author_url": "", "post_date": "09/16/2020 09:45:43", "content": "it is unclear from the challenge 'overview' tab, but as it is a code competition, I guess your code will be re-run against unseen grids to compute the final private leaderboard.
\n@inversion or someone from kaggle team : can you clarify this point?
", "votes": null, "replies": [ { "id": 1013798, "author_name": "jamesmcguigan", "author_url": "", "post_date": "09/17/2020 01:49:48", "content": "I believe in this competition the leaderboard data is the same as the public test dataset.
\nIt would be nice if somebody on the Kaggle team could confirm this, and if this applies to both the public and the private leaderboards.
", "votes": null, "replies": [] }, { "id": 1062228, "author_name": "ajbradberry96", "author_url": "", "post_date": "10/27/2020 17:13:06", "content": "Yes, and it seems as if the leaderboard data that is being displayed now is only for half of the grids that are in the test set, is this correct?
", "votes": null, "replies": [] } ] }, { "id": 1013228, "author_name": "inversion", "author_url": "", "post_date": "09/16/2020 15:17:16", "content": "@olegtrott Great question, and it can be confusing with the different formats we have.
\nThis is a Code competition in the sense that a Notebook must be submitted. This is the only way that we can trigger the Game of Life solver on your solution.
\nBut your notebook can be anything from: 1) full training and inference, to 2) just inference from pretrained weights, and even just 3) a Notebook that reads a csv submission from a private dataset and submits that.
\nIn other words, you can do everything offline, even the predictions, and then have the Notebook just read/submit that prediction file.
\nHave fun and good luck!
", "votes": null, "replies": [ { "id": 1045987, "author_name": "nosferican", "author_url": "", "post_date": "10/11/2020 08:12:34", "content": "I am having some issues getting that to work. I have uploaded my submission.csv file and tried triggering the kernel for registering it as an output but keep getting an error. My notebook has
\ntbl <- read.csv(file = file.path(\"..\", \"input\", \"rivanna\", \"data\", \"submission.csv\"))\nwrite.csv(x = tbl, file = \"submission.csv\")\nwrite.csv(x = tbl, file = file.path(\"..\", \"submission.csv\"))\nThe submission.csv shows up after committing the notebook but keep getting the
Submission Scoring Error\nEvaluation Exception: No submission file found. Expected submission.csv\nFollow-up Q: where can we find resource and time constraints for the code execution? (Nevermind, found it: https://www.kaggle.com/c/conways-reverse-game-of-life-2020/overview/code-requirements )
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1012437": "Do the rules prevent players from solving the problem offline and just uploading the results? How?\n\nWhat if they train a NN on the results and upload the weights instead? (In practice, this makes no difference, since NNs tend to memorize the training data)", "1012800": "it is unclear from the challenge 'overview' tab, but as it is a code competition, I guess your code will be re-run against unseen grids to compute the final private leaderboard.\n\n@inversion or someone from kaggle team : can you clarify this point?", "1013228": "olegtrott Great question, and it can be confusing with the different formats we have.\n\nThis is a Code competition in the sense that a Notebook must be submitted. This is the only way that we can trigger the Game of Life solver on your solution.\n\nBut your notebook can be anything from: 1) full training and inference, to 2) just inference from pretrained weights, and even just 3) a Notebook that reads a csv submission from a private dataset and submits that.\n\nIn other words, you can do everything offline, even the predictions, and then have the Notebook just read/submit that prediction file.\n\nHave fun and good luck!", "1013798": "I believe in this competition the leaderboard data is the same as the public test dataset.\n\nIt would be nice if somebody on the Kaggle team could confirm this, and if this applies to both the public and the private leaderboards.", "1028534": "Follow-up Q: where can we find resource and time constraints for the code execution? (Nevermind, found it: https://www.kaggle.com/c/conways-reverse-game-of-life-2020/overview/code-requirements )", "1045987": "I am having some issues getting that to work. I have uploaded my submission.csv file and tried triggering the kernel for registering it as an output but keep getting an error. My notebook has\n```\ntbl <- read.csv(file = file.path(\"..\", \"input\", \"rivanna\", \"data\", \"submission.csv\"))\nwrite.csv(x = tbl, file = \"submission.csv\")\nwrite.csv(x = tbl, file = file.path(\"..\", \"submission.csv\"))\n```\nThe `submission.csv` shows up after committing the notebook but keep getting the\n```\nSubmission Scoring Error\nEvaluation Exception: No submission file found. Expected submission.csv\n```", "1062228": "Yes, and it seems as if the leaderboard data that is being displayed now is only for half of the grids that are in the test set, is this correct?" }, "source": "meta" }