What is best model choice for this kind of problem (time series window detection).
\nI see most public notebook using gradient boosting, XGB, LGBM and randomforest.
\nWhat are some reasons for choosing these models?
Why not use RNN , LSTM or transformer here?
", "messages": [ { "id": "2442487", "postDate": "09/17/2023 04:29:46", "content": "What is best model choice for this kind of problem (time series window detection).
\nI see most public notebook using gradient boosting, XGB, LGBM and randomforest.
\nWhat are some reasons for choosing these models?
Why not use RNN , LSTM or transformer here?
", "rawMarkdown": "What is best model choice for this kind of problem (time series window detection). \nI see most public notebook using gradient boosting, XGB, LGBM and randomforest. \nWhat are some reasons for choosing these models?\n\nWhy not use RNN , LSTM or transformer here?", "votes": null }, { "id": "2442905", "postDate": "09/17/2023 11:43:47", "content": "At this early stage of the competition, I suspect that the popularity of gradient boosted trees is largely a matter of convenience.
\nMost of the public notebooks can be traced back to Carl McBride Ellis' work, which acts as a template for ensemble learning methods. Using his template, teams can dive into the fun part of the competition -- testing new features and models -- without worrying about designing their own pipeline.
\nMeanwhile, I haven't seen any public notebooks using pytorch or tensorflow models. So teams going that route must start from the very beginning: deciding how to frame the problem, creating dataloaders/datasets, trying different architectures, creating a submission.csv from raw outputs, etc.
\nAn XGB model might end up winning the competition, but I'm sure we'll see public notebooks using deep learning as the competition goes on.
", "rawMarkdown": "At this early stage of the competition, I suspect that the popularity of gradient boosted trees is largely a matter of convenience. \n\nMost of the public notebooks can be traced back to Carl McBride Ellis' work, which acts as a template for ensemble learning methods. Using his template, teams can dive into the fun part of the competition -- testing new features and models -- without worrying about designing their own pipeline. \n\nMeanwhile, I haven't seen any public notebooks using pytorch or tensorflow models. So teams going that route must start from the very beginning: deciding how to frame the problem, creating dataloaders/datasets, trying different architectures, creating a submission.csv from raw outputs, etc.\n\nAn XGB model might end up winning the competition, but I'm sure we'll see public notebooks using deep learning as the competition goes on.", "votes": null }, { "id": "2443387", "postDate": "09/17/2023 17:28:21", "content": "Agree with this. I suspect that the top scorers now are using deep learning models (therefore not open-sourcing their solution) and with full data (instead of the Carl's lightweight version) to promote generalizability.
", "rawMarkdown": "Agree with this. I suspect that the top scorers now are using deep learning models (therefore not open-sourcing their solution) and with full data (instead of the Carl's lightweight version) to promote generalizability.", "votes": null }, { "id": "2443475", "postDate": "09/17/2023 18:16:57", "content": "Training with more data and larger models results in a failed submission trying to submit results.
\nThe error type indicates:
\n\"Notebook Threw Exception: While rerunning your code, your notebook hit an unhandled error. Note that the hidden dataset can be larger/smaller/different than the public dataset\"
\nMy code runs from start to finish and generates a submission file but fails during the Kaggle evaluation and submission process. Anyone have this experience recently?
", "rawMarkdown": "Training with more data and larger models results in a failed submission trying to submit results.\n\nThe error type indicates:\n\n\"Notebook Threw Exception: While rerunning your code, your notebook hit an unhandled error. Note that the hidden dataset can be larger/smaller/different than the public dataset\"\n\nMy code runs from start to finish and generates a submission file but fails during the Kaggle evaluation and submission process. Anyone have this experience recently?", "votes": null }, { "id": "2443697", "postDate": "09/17/2023 21:48:48", "content": "How would a time series window detection model architecture look like? I know about RNN and LSTM, and can use them in a model, but struggle to use them with broader utility than predicting the next step in a sequence. Please someone respond I cannot find resources relating to this question.
", "rawMarkdown": "How would a time series window detection model architecture look like? I know about RNN and LSTM, and can use them in a model, but struggle to use them with broader utility than predicting the next step in a sequence. Please someone respond I cannot find resources relating to this question.", "votes": null }, { "id": "2444155", "postDate": "09/18/2023 07:36:58", "content": "Please check submission file is in required format , for me sometimes index column created issues with submission.
", "rawMarkdown": "Please check submission file is in required format , for me sometimes index column created issues with submission.", "votes": null }, { "id": "2444289", "postDate": "09/18/2023 08:40:17", "content": "Hi Rohit,
\nThe submission file error has its own error according to Kaggle Help:
\n\"Submission Scoring Error: Your notebook generated a submission file with incorrect format. Some examples causing this are: wrong number of rows or columns, empty values, an incorrect data type for a value, or invalid submission values from what is expected.\"
\nThey don't really point out what could specifically cause an Exception. This is their take:
\n\"Notebook Threw Exception: While rerunning your code, your notebook hit an unhandled error. Note that the hidden dataset can be larger/smaller/different than the public dataset.\"
\nHowever I figure out the problem was due to some initialization issues, which were not picked up in the notebook execution on my end.
", "rawMarkdown": "Hi Rohit,\n\nThe submission file error has its own error according to Kaggle Help:\n\n\"Submission Scoring Error: Your notebook generated a submission file with incorrect format. Some examples causing this are: wrong number of rows or columns, empty values, an incorrect data type for a value, or invalid submission values from what is expected.\"\n\nThey don't really point out what could specifically cause an Exception. This is their take:\n\n\"Notebook Threw Exception: While rerunning your code, your notebook hit an unhandled error. Note that the hidden dataset can be larger/smaller/different than the public dataset.\"\n\nHowever I figure out the problem was due to some initialization issues, which were not picked up in the notebook execution on my end.", "votes": null }, { "id": "2445886", "postDate": "09/19/2023 06:17:04", "content": "hi, recently there is a similar competition, Parkinson's Freezing of Gait Prediction , in which kagglers used u-net, transformer, wavenet and so on, you can also try them😄.
\nIn addition, the situation of this competition can answer your question. Few people publicly showcase the use of deep learning techniques in their solutions, but that does not mean that no one is using them privately.
", "rawMarkdown": "hi, recently there is a similar competition, [Parkinson's Freezing of Gait Prediction](https://www.kaggle.com/competitions/tlvmc-parkinsons-freezing-gait-prediction) , in which kagglers used u-net, transformer, wavenet and so on, you can also try them😄.\n\nIn addition, the situation of this competition can answer your question. Few people publicly showcase the use of deep learning techniques in their solutions, but that does not mean that no one is using them privately.", "votes": null } ], "comments": [ { "id": 2442905, "author_name": "quillen", "author_url": "", "post_date": "09/17/2023 11:43:47", "content": "At this early stage of the competition, I suspect that the popularity of gradient boosted trees is largely a matter of convenience.
\nMost of the public notebooks can be traced back to Carl McBride Ellis' work, which acts as a template for ensemble learning methods. Using his template, teams can dive into the fun part of the competition -- testing new features and models -- without worrying about designing their own pipeline.
\nMeanwhile, I haven't seen any public notebooks using pytorch or tensorflow models. So teams going that route must start from the very beginning: deciding how to frame the problem, creating dataloaders/datasets, trying different architectures, creating a submission.csv from raw outputs, etc.
\nAn XGB model might end up winning the competition, but I'm sure we'll see public notebooks using deep learning as the competition goes on.
", "votes": null, "replies": [ { "id": 2443387, "author_name": "rimbax", "author_url": "", "post_date": "09/17/2023 17:28:21", "content": "Agree with this. I suspect that the top scorers now are using deep learning models (therefore not open-sourcing their solution) and with full data (instead of the Carl's lightweight version) to promote generalizability.
", "votes": null, "replies": [] } ] }, { "id": 2443475, "author_name": "aidlre001", "author_url": "", "post_date": "09/17/2023 18:16:57", "content": "Training with more data and larger models results in a failed submission trying to submit results.
\nThe error type indicates:
\n\"Notebook Threw Exception: While rerunning your code, your notebook hit an unhandled error. Note that the hidden dataset can be larger/smaller/different than the public dataset\"
\nMy code runs from start to finish and generates a submission file but fails during the Kaggle evaluation and submission process. Anyone have this experience recently?
", "votes": null, "replies": [ { "id": 2444155, "author_name": "rohitchaudhari25", "author_url": "", "post_date": "09/18/2023 07:36:58", "content": "Please check submission file is in required format , for me sometimes index column created issues with submission.
", "votes": null, "replies": [ { "id": 2444289, "author_name": "aidlre001", "author_url": "", "post_date": "09/18/2023 08:40:17", "content": "Hi Rohit,
\nThe submission file error has its own error according to Kaggle Help:
\n\"Submission Scoring Error: Your notebook generated a submission file with incorrect format. Some examples causing this are: wrong number of rows or columns, empty values, an incorrect data type for a value, or invalid submission values from what is expected.\"
\nThey don't really point out what could specifically cause an Exception. This is their take:
\n\"Notebook Threw Exception: While rerunning your code, your notebook hit an unhandled error. Note that the hidden dataset can be larger/smaller/different than the public dataset.\"
\nHowever I figure out the problem was due to some initialization issues, which were not picked up in the notebook execution on my end.
", "votes": null, "replies": [] } ] } ] }, { "id": 2443697, "author_name": "aleksanderklimczyk", "author_url": "", "post_date": "09/17/2023 21:48:48", "content": "How would a time series window detection model architecture look like? I know about RNN and LSTM, and can use them in a model, but struggle to use them with broader utility than predicting the next step in a sequence. Please someone respond I cannot find resources relating to this question.
", "votes": null, "replies": [] }, { "id": 2445886, "author_name": "roger92", "author_url": "", "post_date": "09/19/2023 06:17:04", "content": "hi, recently there is a similar competition, Parkinson's Freezing of Gait Prediction , in which kagglers used u-net, transformer, wavenet and so on, you can also try them😄.
\nIn addition, the situation of this competition can answer your question. Few people publicly showcase the use of deep learning techniques in their solutions, but that does not mean that no one is using them privately.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "2442487": "What is best model choice for this kind of problem (time series window detection). \nI see most public notebook using gradient boosting, XGB, LGBM and randomforest. \nWhat are some reasons for choosing these models?\n\nWhy not use RNN , LSTM or transformer here?", "2442905": "At this early stage of the competition, I suspect that the popularity of gradient boosted trees is largely a matter of convenience. \n\nMost of the public notebooks can be traced back to Carl McBride Ellis' work, which acts as a template for ensemble learning methods. Using his template, teams can dive into the fun part of the competition -- testing new features and models -- without worrying about designing their own pipeline. \n\nMeanwhile, I haven't seen any public notebooks using pytorch or tensorflow models. So teams going that route must start from the very beginning: deciding how to frame the problem, creating dataloaders/datasets, trying different architectures, creating a submission.csv from raw outputs, etc.\n\nAn XGB model might end up winning the competition, but I'm sure we'll see public notebooks using deep learning as the competition goes on.", "2443387": "Agree with this. I suspect that the top scorers now are using deep learning models (therefore not open-sourcing their solution) and with full data (instead of the Carl's lightweight version) to promote generalizability.", "2443475": "Training with more data and larger models results in a failed submission trying to submit results.\n\nThe error type indicates:\n\n\"Notebook Threw Exception: While rerunning your code, your notebook hit an unhandled error. Note that the hidden dataset can be larger/smaller/different than the public dataset\"\n\nMy code runs from start to finish and generates a submission file but fails during the Kaggle evaluation and submission process. Anyone have this experience recently?", "2443697": "How would a time series window detection model architecture look like? I know about RNN and LSTM, and can use them in a model, but struggle to use them with broader utility than predicting the next step in a sequence. Please someone respond I cannot find resources relating to this question.", "2444155": "Please check submission file is in required format , for me sometimes index column created issues with submission.", "2444289": "Hi Rohit,\n\nThe submission file error has its own error according to Kaggle Help:\n\n\"Submission Scoring Error: Your notebook generated a submission file with incorrect format. Some examples causing this are: wrong number of rows or columns, empty values, an incorrect data type for a value, or invalid submission values from what is expected.\"\n\nThey don't really point out what could specifically cause an Exception. This is their take:\n\n\"Notebook Threw Exception: While rerunning your code, your notebook hit an unhandled error. Note that the hidden dataset can be larger/smaller/different than the public dataset.\"\n\nHowever I figure out the problem was due to some initialization issues, which were not picked up in the notebook execution on my end.", "2445886": "hi, recently there is a similar competition, [Parkinson's Freezing of Gait Prediction](https://www.kaggle.com/competitions/tlvmc-parkinsons-freezing-gait-prediction) , in which kagglers used u-net, transformer, wavenet and so on, you can also try them😄.\n\nIn addition, the situation of this competition can answer your question. Few people publicly showcase the use of deep learning techniques in their solutions, but that does not mean that no one is using them privately." }, "source": "meta" }