I started to consider time series to use fast sampling/smooting and acoustic functions analysis as filters and harmonic frequency studies. I found this method quite useful until there. \nI open this topic to let discuss about the consideration of using time series ?
", "messages": [ { "id": 457688, "postDate": "2019-01-17T23:59:48.837Z", "content": "I started to consider time series to use fast sampling/smooting and acoustic functions analysis as filters and harmonic frequency studies. I found this method quite useful until there. \nI open this topic to let discuss about the consideration of using time series ?
", "rawMarkdown": "I started to consider time series to use fast sampling/smooting and acoustic functions analysis as filters and harmonic frequency studies. I found this method quite useful until there. \nI open this topic to let discuss about the consideration of using time series ?", "votes": 17 }, { "id": 457847, "postDate": "2019-01-18T07:48:31.350Z", "content": "hi my friend, could you provide an example of \"acoustic functions\", i tried to google it, but find nothing
", "rawMarkdown": "hi my friend, could you provide an example of \"acoustic functions\", i tried to google it, but find nothing", "votes": 3, "replies": [ { "id": 457888, "postDate": "2019-01-18T09:16:47.517Z", "content": "This is my kernel where i used analytics functions for analysis acoustic signal. May be this will helpful for you https://www.kaggle.com/nikitagribov/analysis-function-for-seismic-signal-data
", "rawMarkdown": "This is my kernel where i used analytics functions for analysis acoustic signal. May be this will helpful for you https://www.kaggle.com/nikitagribov/analysis-function-for-seismic-signal-data", "votes": 2 }, { "id": 457890, "postDate": "2019-01-18T09:30:15.747Z", "rawMarkdown": "", "votes": -1, "isDeleted": true }, { "id": 458115, "postDate": "2019-01-18T20:14:54.503Z", "content": "so, hilbert is one of the \"acoustic functions\"?
", "rawMarkdown": "so, hilbert is one of the \"acoustic functions\"?", "votes": 1 }, { "id": 458430, "postDate": "2019-01-19T16:02:20.673Z", "content": "If deep learning can figure out any function mapping from the input to the output, why do you bother doing it yourself?
", "rawMarkdown": "If deep learning can figure out any function mapping from the input to the output, why do you bother doing it yourself?", "votes": -1 }, { "id": 458561, "postDate": "2019-01-19T23:35:57.303Z", "content": "Umm i would argue that \"deep learning can figure out any function\" is not true, at least practically.\nAll models, even linear models can perfectly fit any dataset. \nEach model family has its own favored solution space.
\n\nFor example, in terms of primality test, we haven't seen anyone successfully obtained a neural network that has comparable performance as compared to Miller–Rabin's test. All we can do, at least for now, is to obtain a network that can memorize all the trained prime numbers, which is not good at all.
\n\nAnd there are many more reasons which I would not list here. \nIn short, feature engineering is still necessary; despite of the recent rise of neural network.
", "rawMarkdown": "Umm i would argue that \"deep learning can figure out any function\" is not true, at least practically.\nAll models, even linear models can perfectly fit any dataset. \nEach model family has its own favored solution space.\n\nFor example, in terms of primality test, we haven't seen anyone successfully obtained a neural network that has comparable performance as compared to Miller–Rabin's test. All we can do, at least for now, is to obtain a network that can memorize all the trained prime numbers, which is not good at all.\n\nAnd there are many more reasons which I would not list here. \nIn short, feature engineering is still necessary; despite of the recent rise of neural network.\n", "votes": 1 } ] }, { "id": 457839, "postDate": "2019-01-18T07:11:11.880Z", "rawMarkdown": "", "votes": -2, "isDeleted": true } ], "comments": [ { "id": 457847, "author_name": "Elliot", "author_url": "", "post_date": "2019-01-18T07:48:31.350000", "content": "hi my friend, could you provide an example of \"acoustic functions\", i tried to google it, but find nothing
", "votes": 3, "replies": [ { "id": 457888, "author_name": "Nikita Gribov", "author_url": "", "post_date": "2019-01-18T09:16:47.517000", "content": "This is my kernel where i used analytics functions for analysis acoustic signal. May be this will helpful for you https://www.kaggle.com/nikitagribov/analysis-function-for-seismic-signal-data
", "votes": 2, "replies": [] }, { "id": 457890, "author_name": "", "author_url": "", "post_date": "2019-01-18T09:30:15.747000", "content": "", "votes": -1, "replies": [] }, { "id": 458115, "author_name": "Elliot", "author_url": "", "post_date": "2019-01-18T20:14:54.503000", "content": "so, hilbert is one of the \"acoustic functions\"?
", "votes": 1, "replies": [] }, { "id": 458430, "author_name": "Mousheng Xu", "author_url": "", "post_date": "2019-01-19T16:02:20.673000", "content": "If deep learning can figure out any function mapping from the input to the output, why do you bother doing it yourself?
", "votes": -1, "replies": [] }, { "id": 458561, "author_name": "Elliot", "author_url": "", "post_date": "2019-01-19T23:35:57.303000", "content": "Umm i would argue that \"deep learning can figure out any function\" is not true, at least practically.\nAll models, even linear models can perfectly fit any dataset. \nEach model family has its own favored solution space.
\n\nFor example, in terms of primality test, we haven't seen anyone successfully obtained a neural network that has comparable performance as compared to Miller–Rabin's test. All we can do, at least for now, is to obtain a network that can memorize all the trained prime numbers, which is not good at all.
\n\nAnd there are many more reasons which I would not list here. \nIn short, feature engineering is still necessary; despite of the recent rise of neural network.
", "votes": 1, "replies": [] } ] }, { "id": 457839, "author_name": "", "author_url": "", "post_date": "2019-01-18T07:11:11.880000", "content": "", "votes": -2, "replies": [] } ], "raw_markdown_by_id": { "457688": "I started to consider time series to use fast sampling/smooting and acoustic functions analysis as filters and harmonic frequency studies. I found this method quite useful until there. \nI open this topic to let discuss about the consideration of using time series ?", "457847": "hi my friend, could you provide an example of \"acoustic functions\", i tried to google it, but find nothing", "457839": "" } }