I am very happy that I was able to participate in a dreamy competition with a very interesting theme.
\nHowever I have only one concern.
\nThere is no angle information at all in the currently provided test metadata(test_data.parquet).
\nBecause there is no batch of 661 data corresponding to the test metadata, we don't know which sensor(which geometry point) caught the light.
So, even if a model that can produce good results on test data is born, I think it is more likely that it is a fluke.
\nIf the same input value as the test data is assumed in operation,
\n\"event_id\" and \"batch_id\" is increases over time, independent of direction.
\n\"first_pulse_index\", \"last_pulse_index\" is indicates where the data is in the file.
\nThe model will be an estimation based on these 4 parameters.
(It is a different story if the direction in which the Cherenkov light is emitted has a constant period with time.)
\nWhat are your thoughts on this matter?
\nI just joined kaggle, and this is my first time participating in a competition.
\nIf I said something wrong, I would appreciate it if you could correct me.
Thank you!
", "messages": [ { "id": "2142478", "postDate": "02/13/2023 14:59:56", "content": "I am very happy that I was able to participate in a dreamy competition with a very interesting theme.
\nHowever I have only one concern.
\nThere is no angle information at all in the currently provided test metadata(test_data.parquet).
\nBecause there is no batch of 661 data corresponding to the test metadata, we don't know which sensor(which geometry point) caught the light.
So, even if a model that can produce good results on test data is born, I think it is more likely that it is a fluke.
\nIf the same input value as the test data is assumed in operation,
\n\"event_id\" and \"batch_id\" is increases over time, independent of direction.
\n\"first_pulse_index\", \"last_pulse_index\" is indicates where the data is in the file.
\nThe model will be an estimation based on these 4 parameters.
(It is a different story if the direction in which the Cherenkov light is emitted has a constant period with time.)
\nWhat are your thoughts on this matter?
\nI just joined kaggle, and this is my first time participating in a competition.
\nIf I said something wrong, I would appreciate it if you could correct me.
Thank you!
", "rawMarkdown": "I am very happy that I was able to participate in a dreamy competition with a very interesting theme.\n\nHowever I have only one concern.\nThere is no angle information at all in the currently provided test metadata(test_data.parquet).\nBecause there is no batch of 661 data corresponding to the test metadata, we don't know which sensor(which geometry point) caught the light.\n\nSo, even if a model that can produce good results on test data is born, I think it is more likely that it is a fluke.\n\nIf the same input value as the test data is assumed in operation,\n\"event_id\" and \"batch_id\" is increases over time, independent of direction.\n\"first_pulse_index\", \"last_pulse_index\" is indicates where the data is in the file.\nThe model will be an estimation based on these 4 parameters.\n\n(It is a different story if the direction in which the Cherenkov light is emitted has a constant period with time.)\n\nWhat are your thoughts on this matter?\n\nI just joined kaggle, and this is my first time participating in a competition.\nIf I said something wrong, I would appreciate it if you could correct me.\n\nThank you!", "votes": null }, { "id": "2142845", "postDate": "02/13/2023 20:24:59", "content": "You said
\n\n\nThe model will be an estimation based on these 4 parameters.
\n
which seems to imply that you are using event_id, batch_id, first_pulse_index, and last_pulse_index to predict the azimuth and zenith angles. You will need use the data of the pulses found in the train/batch_.parquet to predict the output. The test data does not provide angles because you are supposed to predict those.
", "rawMarkdown": "You said \n>The model will be an estimation based on these 4 parameters.\n\nwhich seems to imply that you are using event_id, batch_id, first_pulse_index, and last_pulse_index to predict the azimuth and zenith angles. You will need use the data of the pulses found in the train/batch_There is a nice shared notebook on adding features to the meta and sensor data that can be used in a basic model.
", "rawMarkdown": "There is a nice shared [notebook](https://www.kaggle.com/code/utm529fg/eng-featureengineering#5.-Save) on adding features to the meta and sensor data that can be used in a basic model.", "votes": null }, { "id": "2151317", "postDate": "02/20/2023 03:09:01", "content": "Thank you for your comment. It seems that I didn't study enough.
", "rawMarkdown": "Thank you for your comment. It seems that I didn't study enough.", "votes": null }, { "id": "2151318", "postDate": "02/20/2023 03:09:45", "content": "Thank you for your comment. It seems that I didn't study enough. I will refer to the notebook you recommended.
", "rawMarkdown": "Thank you for your comment. It seems that I didn't study enough. I will refer to the notebook you recommended.", "votes": null } ], "comments": [ { "id": 2142845, "author_name": "solverworld", "author_url": "", "post_date": "02/13/2023 20:24:59", "content": "You said
\n\n\nThe model will be an estimation based on these 4 parameters.
\n
which seems to imply that you are using event_id, batch_id, first_pulse_index, and last_pulse_index to predict the azimuth and zenith angles. You will need use the data of the pulses found in the train/batch_.parquet to predict the output. The test data does not provide angles because you are supposed to predict those.
", "votes": null, "replies": [ { "id": 2151317, "author_name": "appleyuki", "author_url": "", "post_date": "02/20/2023 03:09:01", "content": "Thank you for your comment. It seems that I didn't study enough.
", "votes": null, "replies": [] } ] }, { "id": 2143121, "author_name": "pcjimmmy", "author_url": "", "post_date": "02/14/2023 04:04:18", "content": "There is a nice shared notebook on adding features to the meta and sensor data that can be used in a basic model.
", "votes": null, "replies": [ { "id": 2151318, "author_name": "appleyuki", "author_url": "", "post_date": "02/20/2023 03:09:45", "content": "Thank you for your comment. It seems that I didn't study enough. I will refer to the notebook you recommended.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "2142478": "I am very happy that I was able to participate in a dreamy competition with a very interesting theme.\n\nHowever I have only one concern.\nThere is no angle information at all in the currently provided test metadata(test_data.parquet).\nBecause there is no batch of 661 data corresponding to the test metadata, we don't know which sensor(which geometry point) caught the light.\n\nSo, even if a model that can produce good results on test data is born, I think it is more likely that it is a fluke.\n\nIf the same input value as the test data is assumed in operation,\n\"event_id\" and \"batch_id\" is increases over time, independent of direction.\n\"first_pulse_index\", \"last_pulse_index\" is indicates where the data is in the file.\nThe model will be an estimation based on these 4 parameters.\n\n(It is a different story if the direction in which the Cherenkov light is emitted has a constant period with time.)\n\nWhat are your thoughts on this matter?\n\nI just joined kaggle, and this is my first time participating in a competition.\nIf I said something wrong, I would appreciate it if you could correct me.\n\nThank you!", "2142845": "You said \n>The model will be an estimation based on these 4 parameters.\n\nwhich seems to imply that you are using event_id, batch_id, first_pulse_index, and last_pulse_index to predict the azimuth and zenith angles. You will need use the data of the pulses found in the train/batch_