Hi everyone,\ndid anybody played around with der scipy welch function? I used it and i got an instand result of 1.5-1.6 just with 128 welch features (default output of the function without parameters). Did anybody tried them as well and has some ideas on how to optimize them?
", "messages": [ { "id": 537598, "postDate": "2019-05-27T09:46:30.443Z", "content": "Hi everyone,\ndid anybody played around with der scipy welch function? I used it and i got an instand result of 1.5-1.6 just with 128 welch features (default output of the function without parameters). Did anybody tried them as well and has some ideas on how to optimize them?
", "rawMarkdown": "Hi everyone,\ndid anybody played around with der scipy welch function? I used it and i got an instand result of 1.5-1.6 just with 128 welch features (default output of the function without parameters). Did anybody tried them as well and has some ideas on how to optimize them?", "votes": 3 }, { "id": 537612, "postDate": "2019-05-27T10:26:27.587Z", "content": "its possible to get better than 1.5 using just statistical features such as mean, median , std, and kurt etc
", "rawMarkdown": "its possible to get better than 1.5 using just statistical features such as mean, median , std, and kurt etc", "replies": [ { "id": 537617, "postDate": "2019-05-27T10:42:05.107Z", "content": "Off course they're important. I use then as well. That's why i have a better score then 1.5. \nMy question is if there are maybe some interesting features which can be extracted with the welch function. And if someone already has some results with it.
", "rawMarkdown": "Off course they're important. I use then as well. That's why i have a better score then 1.5. \nMy question is if there are maybe some interesting features which can be extracted with the welch function. And if someone already has some results with it. " }, { "id": 537618, "postDate": "2019-05-27T10:45:05.920Z", "content": "Top people has best scores in less than 20 features. Well we will know in a week.
", "rawMarkdown": "Top people has best scores in less than 20 features. Well we will know in a week." }, { "id": 537620, "postDate": "2019-05-27T10:54:46.317Z", "content": "We don't know who top people are for now ;) Public LB is very misleading IMHO.
", "rawMarkdown": "We don't know who top people are for now ;) Public LB is very misleading IMHO.\n", "votes": 5 }, { "id": 537642, "postDate": "2019-05-27T11:40:19.483Z", "content": "Manoj, i think we talk past one another. By telling i added 128 welch features it wasn't a suggestion to do it like that and it is a way to score high. I just wanted to share my result when play around with different methods, which are used in other projects to process signals, and ask if anybody else had some experience with the function. The welch features always end up taking nearly 50% of my top 20 when calculating the feature importance. So they must have some direct correlation to the result.\nMaybe someone has an idea to process the result of the function and produce some 1-3 features out of it. i Just want to share my experience ...
", "rawMarkdown": "Manoj, i think we talk past one another. By telling i added 128 welch features it wasn't a suggestion to do it like that and it is a way to score high. I just wanted to share my result when play around with different methods, which are used in other projects to process signals, and ask if anybody else had some experience with the function. The welch features always end up taking nearly 50% of my top 20 when calculating the feature importance. So they must have some direct correlation to the result.\nMaybe someone has an idea to process the result of the function and produce some 1-3 features out of it. i Just want to share my experience ...", "votes": 1 }, { "id": 542855, "postDate": "2019-06-04T06:36:27.040Z", "content": "Made it to 124th with the welch approach. Should have stayed on that path and optimized it. Next time i listen more at my self. Lesson learned.
", "rawMarkdown": "Made it to 124th with the welch approach. Should have stayed on that path and optimized it. Next time i listen more at my self. Lesson learned. ", "votes": 1 } ] }, { "id": 537765, "postDate": "2019-05-27T15:18:13.873Z", "rawMarkdown": "", "votes": -4, "isDeleted": true } ], "comments": [ { "id": 537612, "author_name": "Manoj", "author_url": "", "post_date": "2019-05-27T10:26:27.587000", "content": "its possible to get better than 1.5 using just statistical features such as mean, median , std, and kurt etc
", "votes": 0, "replies": [ { "id": 537617, "author_name": "SJ", "author_url": "", "post_date": "2019-05-27T10:42:05.107000", "content": "Off course they're important. I use then as well. That's why i have a better score then 1.5. \nMy question is if there are maybe some interesting features which can be extracted with the welch function. And if someone already has some results with it.
", "votes": 0, "replies": [] }, { "id": 537618, "author_name": "Manoj", "author_url": "", "post_date": "2019-05-27T10:45:05.920000", "content": "Top people has best scores in less than 20 features. Well we will know in a week.
", "votes": 0, "replies": [] }, { "id": 537620, "author_name": "CPMP", "author_url": "", "post_date": "2019-05-27T10:54:46.317000", "content": "We don't know who top people are for now ;) Public LB is very misleading IMHO.
", "votes": 5, "replies": [] }, { "id": 537642, "author_name": "SJ", "author_url": "", "post_date": "2019-05-27T11:40:19.483000", "content": "Manoj, i think we talk past one another. By telling i added 128 welch features it wasn't a suggestion to do it like that and it is a way to score high. I just wanted to share my result when play around with different methods, which are used in other projects to process signals, and ask if anybody else had some experience with the function. The welch features always end up taking nearly 50% of my top 20 when calculating the feature importance. So they must have some direct correlation to the result.\nMaybe someone has an idea to process the result of the function and produce some 1-3 features out of it. i Just want to share my experience ...
", "votes": 1, "replies": [] }, { "id": 542855, "author_name": "SJ", "author_url": "", "post_date": "2019-06-04T06:36:27.040000", "content": "Made it to 124th with the welch approach. Should have stayed on that path and optimized it. Next time i listen more at my self. Lesson learned.
", "votes": 1, "replies": [] } ] }, { "id": 537765, "author_name": "", "author_url": "", "post_date": "2019-05-27T15:18:13.873000", "content": "", "votes": -4, "replies": [] } ], "raw_markdown_by_id": { "537598": "Hi everyone,\ndid anybody played around with der scipy welch function? I used it and i got an instand result of 1.5-1.6 just with 128 welch features (default output of the function without parameters). Did anybody tried them as well and has some ideas on how to optimize them?", "537612": "its possible to get better than 1.5 using just statistical features such as mean, median , std, and kurt etc", "537765": "" } }