{ "id": 443330, "title": "Question about SMILES in the test dataset", "url": "/competitions/open-problems-single-cell-perturbations/discussion/443330", "author_name": "", "post_date": "2023-09-26T15:42:15.392520500Z", "votes": 1, "comment_count": 5, "views": 0, "content": "

I was thinking about using SMILES for prediction, but there doesn't seem to be a lookup table for sm_name -> SMILES, other than in the de_train file. Does this mean that for the unseen test dataset, all of the sm_name values will also come from de_train? That way we can do a lookup, apply some function, and add this as a feature to the original train/test features that our ML model is based on. Is this a correct understanding?

", "messages": [ { "id": "2457076", "postDate": "09/26/2023 15:42:15", "content": "

I was thinking about using SMILES for prediction, but there doesn't seem to be a lookup table for sm_name -> SMILES, other than in the de_train file. Does this mean that for the unseen test dataset, all of the sm_name values will also come from de_train? That way we can do a lookup, apply some function, and add this as a feature to the original train/test features that our ML model is based on. Is this a correct understanding?

", "rawMarkdown": "I was thinking about using SMILES for prediction, but there doesn't seem to be a lookup table for `sm_name` -> `SMILES`, other than in the `de_train` file. Does this mean that for the unseen test dataset, all of the `sm_name` values will also come from `de_train`? That way we can do a lookup, apply some function, and add this as a feature to the original train/test features that our ML model is based on. Is this a correct understanding?", "votes": null }, { "id": "2457125", "postDate": "09/26/2023 16:27:53", "content": "

Or am I just horribly confused? I'm admittedly quite a noob when it comes to Kaggle competitions. Is the id_map the entirety of the test data, and the leaderboard only shows a % of it? Sorry in advance.

", "rawMarkdown": "Or am I just horribly confused? I'm admittedly quite a noob when it comes to Kaggle competitions. Is the `id_map` the entirety of the test data, and the leaderboard only shows a % of it? Sorry in advance.", "votes": null }, { "id": "2457512", "postDate": "09/27/2023 02:16:19", "content": "
    \n
  1. The de_train.csv contains all the SMILES, so we can do a lookup to apply the sm_name -> SMILES for the id_map.csv.
    2. \n
  2. The id_map.csv is the entirety of the test data, but the public leaderboard score only shows a part of it.
    3. \n
  3. You can find these infos in the Data page or my notebook
    4. \n
", "rawMarkdown": "1. The `de_train.csv` contains all the SMILES, so we can do a lookup to apply the `sm_name -> SMILES` for the `id_map.csv`.\n2. The `id_map.csv` is the entirety of the test data, but the public leaderboard score only shows a part of it.\n3. You can find these infos in the [Data page](https://www.kaggle.com/competitions/open-problems-single-cell-perturbations/data) or my [notebook](https://www.kaggle.com/code/awater1223/op2-00-basic-metadata-eda#Data-split)", "votes": null }, { "id": "2457553", "postDate": "09/27/2023 03:26:34", "content": "

Thanks for clarifying!

", "rawMarkdown": "Thanks for clarifying!", "votes": null }, { "id": "2514199", "postDate": "11/06/2023 05:50:14", "content": "

Thanks for your reply, I tried to do so, and I found that only 129 sm_names in id_map are in the de_train.csv

", "rawMarkdown": "Thanks for your reply, I tried to do so, and I found that only 129 sm_names in id_map are in the de_train.csv", "votes": null }, { "id": "2514209", "postDate": "11/06/2023 05:59:45", "content": "

I tried using de_train as a lookup table, but I found that id_map is not entirely lied in de_train, only 129 out of all are in de_train. Have you solve this issue?

", "rawMarkdown": "I tried using de_train as a lookup table, but I found that id_map is not entirely lied in de_train, only 129 out of all are in de_train. Have you solve this issue?", "votes": null } ], "comments": [ { "id": 2457125, "author_name": "yuqizheng", "author_url": "", "post_date": "09/26/2023 16:27:53", "content": "

Or am I just horribly confused? I'm admittedly quite a noob when it comes to Kaggle competitions. Is the id_map the entirety of the test data, and the leaderboard only shows a % of it? Sorry in advance.

", "votes": null, "replies": [ { "id": 2514209, "author_name": "ruiqizhuangricky", "author_url": "", "post_date": "11/06/2023 05:59:45", "content": "

I tried using de_train as a lookup table, but I found that id_map is not entirely lied in de_train, only 129 out of all are in de_train. Have you solve this issue?

", "votes": null, "replies": [] } ] }, { "id": 2457512, "author_name": "awater1223", "author_url": "", "post_date": "09/27/2023 02:16:19", "content": "
    \n
  1. The de_train.csv contains all the SMILES, so we can do a lookup to apply the sm_name -> SMILES for the id_map.csv.
    2. \n
  2. The id_map.csv is the entirety of the test data, but the public leaderboard score only shows a part of it.
    3. \n
  3. You can find these infos in the Data page or my notebook
    4. \n
", "votes": null, "replies": [ { "id": 2457553, "author_name": "yuqizheng", "author_url": "", "post_date": "09/27/2023 03:26:34", "content": "

Thanks for clarifying!

", "votes": null, "replies": [] }, { "id": 2514199, "author_name": "ruiqizhuangricky", "author_url": "", "post_date": "11/06/2023 05:50:14", "content": "

Thanks for your reply, I tried to do so, and I found that only 129 sm_names in id_map are in the de_train.csv

", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "2457076": "I was thinking about using SMILES for prediction, but there doesn't seem to be a lookup table for `sm_name` -> `SMILES`, other than in the `de_train` file. Does this mean that for the unseen test dataset, all of the `sm_name` values will also come from `de_train`? That way we can do a lookup, apply some function, and add this as a feature to the original train/test features that our ML model is based on. Is this a correct understanding?", "2457125": "Or am I just horribly confused? I'm admittedly quite a noob when it comes to Kaggle competitions. Is the `id_map` the entirety of the test data, and the leaderboard only shows a % of it? Sorry in advance.", "2457512": "1. The `de_train.csv` contains all the SMILES, so we can do a lookup to apply the `sm_name -> SMILES` for the `id_map.csv`.\n2. The `id_map.csv` is the entirety of the test data, but the public leaderboard score only shows a part of it.\n3. You can find these infos in the [Data page](https://www.kaggle.com/competitions/open-problems-single-cell-perturbations/data) or my [notebook](https://www.kaggle.com/code/awater1223/op2-00-basic-metadata-eda#Data-split)", "2457553": "Thanks for clarifying!", "2514199": "Thanks for your reply, I tried to do so, and I found that only 129 sm_names in id_map are in the de_train.csv", "2514209": "I tried using de_train as a lookup table, but I found that id_map is not entirely lied in de_train, only 129 out of all are in de_train. Have you solve this issue?" }, "source": "meta" }