At this stage of the competition, most people are computing moving averages, computing differences, and feature crossing to then plug into a machine learning model to predict if the series is currently awake or sleeping. I'm not too interested in this basic approach. Are there deep learning model types or frameworks that are built for the event detection task this competition requires? I am just a beginner so I am unaware of such models. A convolutional type model maybe?
", "messages": [ { "id": "2449071", "postDate": "09/21/2023 02:02:32", "content": "At this stage of the competition, most people are computing moving averages, computing differences, and feature crossing to then plug into a machine learning model to predict if the series is currently awake or sleeping. I'm not too interested in this basic approach. Are there deep learning model types or frameworks that are built for the event detection task this competition requires? I am just a beginner so I am unaware of such models. A convolutional type model maybe?
", "rawMarkdown": "At this stage of the competition, most people are computing moving averages, computing differences, and feature crossing to then plug into a machine learning model to predict if the series is currently awake or sleeping. I'm not too interested in this basic approach. Are there deep learning model types or frameworks that are built for the event detection task this competition requires? I am just a beginner so I am unaware of such models. A convolutional type model maybe?", "votes": null }, { "id": "2449075", "postDate": "09/21/2023 02:08:10", "content": "Since the competition started just recently, everyone is currently using feature engineering and boosted models for their submissions ( as far as we see in the public notebooks ). But I think some have already started using Neural network based models for the competition ( although it's unlikely to see such solutions as a public notebook currently). Maybe some sequential based models like RNNs or LSTMs can be a good place to start while using deep learning models for the competition.
", "rawMarkdown": "Since the competition started just recently, everyone is currently using feature engineering and boosted models for their submissions ( as far as we see in the public notebooks ). But I think some have already started using Neural network based models for the competition ( although it's unlikely to see such solutions as a public notebook currently). Maybe some sequential based models like RNNs or LSTMs can be a good place to start while using deep learning models for the competition.", "votes": null }, { "id": "2449076", "postDate": "09/21/2023 02:08:18", "content": "Hi @aleksanderklimczyk , you can find some discussions regarding the usage of deep learning models in this competition
\nMay it helps.
", "rawMarkdown": "Hi @aleksanderklimczyk , you can find some discussions regarding the usage of deep learning models in this competition\n\n1. [A possible deep learning approach](https://www.kaggle.com/competitions/child-mind-institute-detect-sleep-states/discussion/441470): predicting for every 24h\n2. [Use of Temporal Graph Neural Networks to Detect Sleep States\n](https://www.kaggle.com/competitions/child-mind-institute-detect-sleep-states/discussion/437420): compilation of methods related with TGNN\n3. [🔥Papers with Code for this Competition](https://www.kaggle.com/competitions/child-mind-institute-detect-sleep-states/discussion/437274)\n\nMay it helps.", "votes": null }, { "id": "2449136", "postDate": "09/21/2023 03:11:50", "content": "Thank you.
", "rawMarkdown": "Thank you.", "votes": null }, { "id": "2449372", "postDate": "09/21/2023 06:27:11", "content": "I'm also not very interested in feature engineering, I prefer to use deep learning models. It seems to me that deep learning models automatically learning features should be able to go deeper (in theory) than doing feature engineering manually.
", "rawMarkdown": "I'm also not very interested in feature engineering, I prefer to use deep learning models. It seems to me that deep learning models automatically learning features should be able to go deeper (in theory) than doing feature engineering manually.", "votes": null }, { "id": "2449448", "postDate": "09/21/2023 07:13:08", "content": "Dear @aleksanderklimczyk
\n\"Possible technique? Temporal Convolutional Networks\"
\nAll the best and good luck!
\ncarl
Since the competition started just recently, everyone is currently using feature engineering and boosted models for their submissions ( as far as we see in the public notebooks ). But I think some have already started using Neural network based models for the competition ( although it's unlikely to see such solutions as a public notebook currently). Maybe some sequential based models like RNNs or LSTMs can be a good place to start while using deep learning models for the competition.
", "votes": null, "replies": [] }, { "id": 2449076, "author_name": "rimbax", "author_url": "", "post_date": "09/21/2023 02:08:18", "content": "Hi @aleksanderklimczyk , you can find some discussions regarding the usage of deep learning models in this competition
\nMay it helps.
", "votes": null, "replies": [ { "id": 2449136, "author_name": "aleksanderklimczyk", "author_url": "", "post_date": "09/21/2023 03:11:50", "content": "Thank you.
", "votes": null, "replies": [] } ] }, { "id": 2449372, "author_name": "jimmyisme1", "author_url": "", "post_date": "09/21/2023 06:27:11", "content": "I'm also not very interested in feature engineering, I prefer to use deep learning models. It seems to me that deep learning models automatically learning features should be able to go deeper (in theory) than doing feature engineering manually.
", "votes": null, "replies": [] }, { "id": 2449448, "author_name": "carlmcbrideellis", "author_url": "", "post_date": "09/21/2023 07:13:08", "content": "Dear @aleksanderklimczyk
\n\"Possible technique? Temporal Convolutional Networks\"
\nAll the best and good luck!
\ncarl