What is a batch of predictions? Is it a dataset with the same data_id or time_id? When we make predictions, I think corresponding lags file will be provided. If so, then when we make predictions, the lags file and test file with the same date_id will be provided. Am I right?
", "messages": [ { "id": "3095067", "postDate": "01/12/2025 22:21:12", "content": "What is a batch of predictions? Is it a dataset with the same data_id or time_id? When we make predictions, I think corresponding lags file will be provided. If so, then when we make predictions, the lags file and test file with the same date_id will be provided. Am I right?
", "rawMarkdown": "What is a batch of predictions? Is it a dataset with the same data_id or time_id? When we make predictions, I think corresponding lags file will be provided. If so, then when we make predictions, the lags file and test file with the same date_id will be provided. Am I right?", "votes": null }, { "id": "3095086", "postDate": "01/12/2025 23:34:25", "content": "We get a single time_id at a time for prediction. e.g roughly 30-40 rows per batch (1 row for each symbol_id in the day)
\nWe get lags at time_id=0 only for each day (first batch of each date_id). The lags are for the entire 968 timesteps of the previous day.
\nHere is some sample code you can drop straight into your notebook and run to validate your submission works with lags etc before submitting https://www.kaggle.com/competitions/jane-street-real-time-market-data-forecasting/discussion/555226
", "rawMarkdown": "We get a single time_id at a time for prediction. e.g roughly 30-40 rows per batch (1 row for each symbol_id in the day)\n\nWe get lags at time_id=0 only for each day (first batch of each date_id). The lags are for the entire 968 timesteps of the previous day.\n\nHere is some sample code you can drop straight into your notebook and run to validate your submission works with lags etc before submitting https://www.kaggle.com/competitions/jane-street-real-time-market-data-forecasting/discussion/555226", "votes": null }, { "id": "3095095", "postDate": "01/13/2025 00:23:20", "content": "Many thanks for that. May I just confirm that for each prediction batch (i.e., time_id), lags is not None? In other words, when we do every prediction batch (single time_id), the lags will be provided and the same for one date_id. Am I right?
", "rawMarkdown": "Many thanks for that. May I just confirm that for each prediction batch (i.e., time_id), lags is not None? In other words, when we do every prediction batch (single time_id), the lags will be provided and the same for one date_id. Am I right?", "votes": null }, { "id": "3095106", "postDate": "01/13/2025 00:49:19", "content": "We only get lags at time_id == 0. So if there are 968 batches a day, lags are None for 967 of them. You will need to store the lags in memory if you want to re-use them between time_ids
", "rawMarkdown": "We only get lags at time_id == 0. So if there are 968 batches a day, lags are None for 967 of them. You will need to store the lags in memory if you want to re-use them between time_ids", "votes": null } ], "comments": [ { "id": 3095086, "author_name": "michaeltimbs", "author_url": "", "post_date": "01/12/2025 23:34:25", "content": "We get a single time_id at a time for prediction. e.g roughly 30-40 rows per batch (1 row for each symbol_id in the day)
\nWe get lags at time_id=0 only for each day (first batch of each date_id). The lags are for the entire 968 timesteps of the previous day.
\nHere is some sample code you can drop straight into your notebook and run to validate your submission works with lags etc before submitting https://www.kaggle.com/competitions/jane-street-real-time-market-data-forecasting/discussion/555226
", "votes": null, "replies": [ { "id": 3095095, "author_name": "weiw1422", "author_url": "", "post_date": "01/13/2025 00:23:20", "content": "Many thanks for that. May I just confirm that for each prediction batch (i.e., time_id), lags is not None? In other words, when we do every prediction batch (single time_id), the lags will be provided and the same for one date_id. Am I right?
", "votes": null, "replies": [ { "id": 3095106, "author_name": "michaeltimbs", "author_url": "", "post_date": "01/13/2025 00:49:19", "content": "We only get lags at time_id == 0. So if there are 968 batches a day, lags are None for 967 of them. You will need to store the lags in memory if you want to re-use them between time_ids
", "votes": null, "replies": [] } ] } ] } ], "raw_markdown_by_id": { "3095067": "What is a batch of predictions? Is it a dataset with the same data_id or time_id? When we make predictions, I think corresponding lags file will be provided. If so, then when we make predictions, the lags file and test file with the same date_id will be provided. Am I right?", "3095086": "We get a single time_id at a time for prediction. e.g roughly 30-40 rows per batch (1 row for each symbol_id in the day)\n\nWe get lags at time_id=0 only for each day (first batch of each date_id). The lags are for the entire 968 timesteps of the previous day.\n\nHere is some sample code you can drop straight into your notebook and run to validate your submission works with lags etc before submitting https://www.kaggle.com/competitions/jane-street-real-time-market-data-forecasting/discussion/555226", "3095095": "Many thanks for that. May I just confirm that for each prediction batch (i.e., time_id), lags is not None? In other words, when we do every prediction batch (single time_id), the lags will be provided and the same for one date_id. Am I right?", "3095106": "We only get lags at time_id == 0. So if there are 968 batches a day, lags are None for 967 of them. You will need to store the lags in memory if you want to re-use them between time_ids" }, "source": "meta" }