I'm new at this, but have been reading white papers for a year. Want to use protocol buffers, normalize data, looking at time series stuff closely to try to refine the data as much as possible. Previous experience is a plus, but not required, I'd prefer you be willing to put some quality time in on it, though.
\n\nFrom all appearances (kernels), the time series data outside of the cheaty expire date is not important. This worries me that using the time series data we may not be able to get a neural network to converge (weak signal). But this is for a music streaming service and I want to believe that the user's listening strongly impacts their re-subscription. And these are the technologies I want to learn to use.
\n\nIf you would like to approach this problem from this perspective, let me know!
", "messages": [ { "id": "225851", "postDate": "09/30/2017 05:19:57", "content": "I'm new at this, but have been reading white papers for a year. Want to use protocol buffers, normalize data, looking at time series stuff closely to try to refine the data as much as possible. Previous experience is a plus, but not required, I'd prefer you be willing to put some quality time in on it, though.
\n\nFrom all appearances (kernels), the time series data outside of the cheaty expire date is not important. This worries me that using the time series data we may not be able to get a neural network to converge (weak signal). But this is for a music streaming service and I want to believe that the user's listening strongly impacts their re-subscription. And these are the technologies I want to learn to use.
\n\nIf you would like to approach this problem from this perspective, let me know!
", "rawMarkdown": "I'm new at this, but have been reading white papers for a year. Want to use protocol buffers, normalize data, looking at time series stuff closely to try to refine the data as much as possible. Previous experience is a plus, but not required, I'd prefer you be willing to put some quality time in on it, though.\n\nFrom all appearances (kernels), the time series data outside of the cheaty expire date is not important. This worries me that using the time series data we may not be able to get a neural network to converge (weak signal). But this is for a music streaming service and I want to believe that the user's listening strongly impacts their re-subscription. And these are the technologies I want to learn to use.\n\nIf you would like to approach this problem from this perspective, let me know!", "votes": null }, { "id": "226064", "postDate": "09/30/2017 22:28:37", "content": "Hey Ryan, we have a small team forming right now. Would you like to be apart of our slack channel? We haven't confirmed the team yet, still assembling. It would be great to have you!
", "rawMarkdown": "Hey Ryan, we have a small team forming right now. Would you like to be apart of our slack channel? We haven't confirmed the team yet, still assembling. It would be great to have you!", "votes": null }, { "id": "226387", "postDate": "10/02/2017 05:07:41", "content": "i will try this out ,thanks for illuminating me
", "rawMarkdown": "i will try this out ,thanks for illuminating me", "votes": null }, { "id": "226521", "postDate": "10/02/2017 15:17:02", "content": "Hey sounds good, email me at rbuck1 at iuhealth dot org.
", "rawMarkdown": "Hey sounds good, email me at rbuck1 at iuhealth dot org.", "votes": null }, { "id": "227086", "postDate": "10/03/2017 17:35:47", "content": "Also, I'm considering building a VAE to fill gaps for the user listener data that the LSTM/GRU would use to predict with. When the transaction logs and the user logs do not line up at the end of the interval, there may be user log data missing. I need to analyze the data more and decide whether data is missing there, or just some users did not listen to music during the period ending on 2/28? Better to generate plausible data accurately and take multiple samples than to misinterpret a lack of logs.
\n\nWe could use statistical techniques to generate realistic distributions, but they would not be customized per the users' other information.
", "rawMarkdown": "Also, I'm considering building a VAE to fill gaps for the user listener data that the LSTM/GRU would use to predict with. *When the transaction logs and the user logs do not line up at the end of the interval, there may be user log data missing.* I need to analyze the data more and decide whether data is missing there, or just some users did not listen to music during the period ending on 2/28? Better to generate plausible data accurately and take multiple samples than to misinterpret a lack of logs.\n\nWe could use statistical techniques to generate realistic distributions, but they would not be customized per the users' other information.", "votes": null } ], "comments": [ { "id": 226064, "author_name": "randylaosat", "author_url": "", "post_date": "09/30/2017 22:28:37", "content": "Hey Ryan, we have a small team forming right now. Would you like to be apart of our slack channel? We haven't confirmed the team yet, still assembling. It would be great to have you!
", "votes": null, "replies": [ { "id": 226521, "author_name": "stravinsky", "author_url": "", "post_date": "10/02/2017 15:17:02", "content": "Hey sounds good, email me at rbuck1 at iuhealth dot org.
", "votes": null, "replies": [] }, { "id": 227086, "author_name": "stravinsky", "author_url": "", "post_date": "10/03/2017 17:35:47", "content": "Also, I'm considering building a VAE to fill gaps for the user listener data that the LSTM/GRU would use to predict with. When the transaction logs and the user logs do not line up at the end of the interval, there may be user log data missing. I need to analyze the data more and decide whether data is missing there, or just some users did not listen to music during the period ending on 2/28? Better to generate plausible data accurately and take multiple samples than to misinterpret a lack of logs.
\n\nWe could use statistical techniques to generate realistic distributions, but they would not be customized per the users' other information.
", "votes": null, "replies": [] } ] }, { "id": 226387, "author_name": "liulongxiao1", "author_url": "", "post_date": "10/02/2017 05:07:41", "content": "i will try this out ,thanks for illuminating me
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "225851": "I'm new at this, but have been reading white papers for a year. Want to use protocol buffers, normalize data, looking at time series stuff closely to try to refine the data as much as possible. Previous experience is a plus, but not required, I'd prefer you be willing to put some quality time in on it, though.\n\nFrom all appearances (kernels), the time series data outside of the cheaty expire date is not important. This worries me that using the time series data we may not be able to get a neural network to converge (weak signal). But this is for a music streaming service and I want to believe that the user's listening strongly impacts their re-subscription. And these are the technologies I want to learn to use.\n\nIf you would like to approach this problem from this perspective, let me know!", "226064": "Hey Ryan, we have a small team forming right now. Would you like to be apart of our slack channel? We haven't confirmed the team yet, still assembling. It would be great to have you!", "226387": "i will try this out ,thanks for illuminating me", "226521": "Hey sounds good, email me at rbuck1 at iuhealth dot org.", "227086": "Also, I'm considering building a VAE to fill gaps for the user listener data that the LSTM/GRU would use to predict with. *When the transaction logs and the user logs do not line up at the end of the interval, there may be user log data missing.* I need to analyze the data more and decide whether data is missing there, or just some users did not listen to music during the period ending on 2/28? Better to generate plausible data accurately and take multiple samples than to misinterpret a lack of logs.\n\nWe could use statistical techniques to generate realistic distributions, but they would not be customized per the users' other information." }, "source": "meta" }