I have been trying to fit pure signal processing techniques in this competition using Scipy's hilbert function to fit square waves to the enmo reading and fit thresholds around the onset/awake periods but have had little success in doing so, I am open-sourcing my notebook here to see what the community might be able to extend from this!
\nEnjoy, let me know any thoughts.
", "messages": [ { "id": "2442217", "postDate": "09/16/2023 20:01:03", "content": "I have been trying to fit pure signal processing techniques in this competition using Scipy's hilbert function to fit square waves to the enmo reading and fit thresholds around the onset/awake periods but have had little success in doing so, I am open-sourcing my notebook here to see what the community might be able to extend from this!
\nEnjoy, let me know any thoughts.
", "rawMarkdown": "I have been trying to fit pure signal processing techniques in this competition using Scipy's hilbert function to fit square waves to the enmo reading and fit thresholds around the onset/awake periods but have had little success in doing so, I am open-sourcing my notebook [here](https://www.kaggle.com/code/munumbutt/signal-processing-techniques?scriptVersionId=143239775) to see what the community might be able to extend from this!\n\nEnjoy, let me know any thoughts.", "votes": null }, { "id": "2442298", "postDate": "09/16/2023 21:57:22", "content": "It is reasonable to try \"signal processing\" techniques, if only to compare them with the more-ML techniques. Your notebook is using the test data which, of course, it should for submission. But have you tried it on the train data to see what it does and how it performs on some 'real' series?
", "rawMarkdown": "It is reasonable to try \"signal processing\" techniques, if only to compare them with the more-ML techniques. Your notebook is using the test data which, of course, it should for submission. But have you tried it on the train data to see what it does and how it performs on some 'real' series?", "votes": null }, { "id": "2444569", "postDate": "09/18/2023 10:46:22", "content": "@munumbutt : Hey Munum, thanks for sharing your contribution.
\nML techniques (and thus trackling this problem as a regression problem) could bring more fruiteful results in the long run.
\nAs per the pure singal processing paradigm, you should try to look at both enmo and anglez in parallel. As demonstrated in Plotting Big Data: Sleeping Series, there is a complex variability pattern around event points that we need to crack through.
\nGood luck in your futher competition contributions!
", "rawMarkdown": "munumbutt : Hey Munum, thanks for sharing your contribution.\n\nML techniques (and thus trackling this problem as a regression problem) could bring more fruiteful results in the long run.\n\nAs per the pure singal processing paradigm, you should try to look at both *enmo* and *anglez* in parallel. As demonstrated in [Plotting Big Data: Sleeping Series](https://www.kaggle.com/code/gvyshnya/plotting-big-data-sleeping-series), there is a complex variability pattern around event points that we need to crack through.\n\nGood luck in your futher competition contributions!", "votes": null } ], "comments": [ { "id": 2442298, "author_name": "dan3dewey", "author_url": "", "post_date": "09/16/2023 21:57:22", "content": "It is reasonable to try \"signal processing\" techniques, if only to compare them with the more-ML techniques. Your notebook is using the test data which, of course, it should for submission. But have you tried it on the train data to see what it does and how it performs on some 'real' series?
", "votes": null, "replies": [] }, { "id": 2444569, "author_name": "gvyshnya", "author_url": "", "post_date": "09/18/2023 10:46:22", "content": "@munumbutt : Hey Munum, thanks for sharing your contribution.
\nML techniques (and thus trackling this problem as a regression problem) could bring more fruiteful results in the long run.
\nAs per the pure singal processing paradigm, you should try to look at both enmo and anglez in parallel. As demonstrated in Plotting Big Data: Sleeping Series, there is a complex variability pattern around event points that we need to crack through.
\nGood luck in your futher competition contributions!
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "2442217": "I have been trying to fit pure signal processing techniques in this competition using Scipy's hilbert function to fit square waves to the enmo reading and fit thresholds around the onset/awake periods but have had little success in doing so, I am open-sourcing my notebook [here](https://www.kaggle.com/code/munumbutt/signal-processing-techniques?scriptVersionId=143239775) to see what the community might be able to extend from this!\n\nEnjoy, let me know any thoughts.", "2442298": "It is reasonable to try \"signal processing\" techniques, if only to compare them with the more-ML techniques. Your notebook is using the test data which, of course, it should for submission. But have you tried it on the train data to see what it does and how it performs on some 'real' series?", "2444569": "munumbutt : Hey Munum, thanks for sharing your contribution.\n\nML techniques (and thus trackling this problem as a regression problem) could bring more fruiteful results in the long run.\n\nAs per the pure singal processing paradigm, you should try to look at both *enmo* and *anglez* in parallel. As demonstrated in [Plotting Big Data: Sleeping Series](https://www.kaggle.com/code/gvyshnya/plotting-big-data-sleeping-series), there is a complex variability pattern around event points that we need to crack through.\n\nGood luck in your futher competition contributions!" }, "source": "meta" }