There isn't a lot of time left to try out new ideas, and there's still a lot of tuning to do on the old ideas. But maybe your old ideas are a little different from mine, and they will be useful to combine? In any case, as far as tuning goes, two computers are better than one, and N*K computers are better than K (where K is the average number of computers per teammate and N is the number of teammates).
\n\nWhat I have now is mostly based on close variations of the frameworks and feature sets in public kernels, with some tuning and swapping of methods applied (and with some attempts to improve validation and to test robustness with respect to the public/private split, and with some mostly unsuccessful testing of additional features). So I'd be particularly interested in teammates that have had some success with other feature engineering or with frameworks other than the one popularized by Ceshine Lee.
\n\nAnyhow, if you have reason to believe we can be a whole that is greater than the sum of its parts, let me know.
", "messages": [ { "id": "264739", "postDate": "01/03/2018 19:36:43", "content": "There isn't a lot of time left to try out new ideas, and there's still a lot of tuning to do on the old ideas. But maybe your old ideas are a little different from mine, and they will be useful to combine? In any case, as far as tuning goes, two computers are better than one, and N*K computers are better than K (where K is the average number of computers per teammate and N is the number of teammates).
\n\nWhat I have now is mostly based on close variations of the frameworks and feature sets in public kernels, with some tuning and swapping of methods applied (and with some attempts to improve validation and to test robustness with respect to the public/private split, and with some mostly unsuccessful testing of additional features). So I'd be particularly interested in teammates that have had some success with other feature engineering or with frameworks other than the one popularized by Ceshine Lee.
\n\nAnyhow, if you have reason to believe we can be a whole that is greater than the sum of its parts, let me know.
", "rawMarkdown": "There isn't a lot of time left to try out new ideas, and there's still a lot of tuning to do on the old ideas. But maybe your old ideas are a little different from mine, and they will be useful to combine? In any case, as far as tuning goes, two computers are better than one, and N*K computers are better than K (where K is the average number of computers per teammate and N is the number of teammates).\n\nWhat I have now is mostly based on close variations of the frameworks and feature sets in public kernels, with some tuning and swapping of methods applied (and with some attempts to improve validation and to test robustness with respect to the public/private split, and with some mostly unsuccessful testing of additional features). So I'd be particularly interested in teammates that have had some success with other feature engineering or with frameworks other than the one popularized by Ceshine Lee.\n\nAnyhow, if you have reason to believe we can be a whole that is greater than the sum of its parts, let me know.", "votes": null }, { "id": "264769", "postDate": "01/03/2018 20:35:22", "content": "Hi Andy, I have achieved a score of 0.512 with different models tried out. I have few more models to try and check. It will be great if we can form a team and share our ideas with each other.
\n\nLet me know.
\n\nThanks.
", "rawMarkdown": "Hi Andy, I have achieved a score of 0.512 with different models tried out. I have few more models to try and check. It will be great if we can form a team and share our ideas with each other. \n\nLet me know. \n\nThanks.", "votes": null }, { "id": "264891", "postDate": "01/04/2018 03:38:57", "content": "Hey Andy. I may be interested, or if there is anyone else out there interested in talking. I don't want to get into detail publicly, but I have a .510 using the LGBM framework. I work in R and use data.table. Models use essentially the same framework as LGBM Starter.
\n\nI don't have a team friendly coding framework, however, I think sharing FE ideas, and some 'novel' framework ideas could be very beneficial.
\n\nI will be able to reply fast and share ideas quickly through the final submission.
\n\nEdit: Found a team.
", "rawMarkdown": "Hey Andy. I may be interested, or if there is anyone else out there interested in talking. I don't want to get into detail publicly, but I have a .510 using the LGBM framework. I work in R and use data.table. Models use essentially the same framework as LGBM Starter. \n \nI don't have a team friendly coding framework, however, I think sharing FE ideas, and some 'novel' framework ideas could be very beneficial. \n \nI will be able to reply fast and share ideas quickly through the final submission.\n\nEdit: Found a team.", "votes": null }, { "id": "265130", "postDate": "01/04/2018 17:27:31", "content": "People think I'm always trolling them, and they're probably right. If you can deal with that would you like to join our team?
", "rawMarkdown": "People think I'm always trolling them, and they're probably right. If you can deal with that would you like to join our team?", "votes": null }, { "id": "265147", "postDate": "01/04/2018 18:09:06", "content": "Hi Andy,\nIt would be an honor to work with you on this competition. If you haven't found a team, yet, then please consider Steven's invite. He's actually a very cordial person - I haven't been trolled, yet - and he's made my on-boarding to the team pleasant and smooth. We are using a couple of services to help with communicating ideas. The teammates are all very respectful, positive and hardworking. We hope you will consider our invite.
\n\nThank you.
", "rawMarkdown": "Hi Andy,\nIt would be an honor to work with you on this competition. If you haven't found a team, yet, then please consider Steven's invite. He's actually a very cordial person - I haven't been trolled, yet - and he's made my on-boarding to the team pleasant and smooth. We are using a couple of services to help with communicating ideas. The teammates are all very respectful, positive and hardworking. We hope you will consider our invite. \n\nThank you.", "votes": null }, { "id": "265158", "postDate": "01/04/2018 18:52:46", "content": "Yes, I've worked with Steven before (on Wikipedia -- and unofficially in public on Instacart, where teaming wasn't allowed). This does seem like a promising possibility. I already have an email reply out to RB (currently #63), who was the first to respond to my post. (I tried to send her a merge request, but it didn't work: seems there is a bug with very short team names.)
\n\nI am particularly interested if anyone has a method for predicting sales of new items. What I have tried so far hasn't worked at all, except for just using a small constant, but I haven't yet tried the obvious things that require a little more database work.
", "rawMarkdown": "Yes, I've worked with Steven before (on Wikipedia -- and unofficially in public on Instacart, where teaming wasn't allowed). This does seem like a promising possibility. I already have an email reply out to RB (currently #63), who was the first to respond to my post. (I tried to send her a merge request, but it didn't work: seems there is a bug with very short team names.)\n\nI am particularly interested if anyone has a method for predicting sales of new items. What I have tried so far hasn't worked at all, except for just using a small constant, but I haven't yet tried the obvious things that require a little more database work.", "votes": null }, { "id": "265164", "postDate": "01/04/2018 19:04:10", "content": "OK, RB has changed her name to \"Rashmi B\", which works with merge requests. Also she sent me a merge request. What do you guys think about a team of 6?
", "rawMarkdown": "OK, RB has changed her name to \"Rashmi B\", which works with merge requests. Also she sent me a merge request. What do you guys think about a team of 6?", "votes": null }, { "id": "265188", "postDate": "01/04/2018 20:21:29", "content": "RB will be a lucky person to have you as a teammate. I'm glad that the issues were corrected. I am open to more the merrier. Let me take this back to the team and get their thoughts. Thank you, again, for keeping open the opportunity to collaborate.
", "rawMarkdown": "RB will be a lucky person to have you as a teammate. I'm glad that the issues were corrected. I am open to more the merrier. Let me take this back to the team and get their thoughts. Thank you, again, for keeping open the opportunity to collaborate.", "votes": null }, { "id": "265202", "postDate": "01/04/2018 21:09:35", "content": "Hi Andy. We'd love to have both you and Rashmi B on the team. Steven was caught in the severe winter storm so we're working on coordinating the invites now. Thank you, again!
\n\nUpdate: seems only 1 invite can go at a time; I will contact Rashmi B to let her know we haven't forgotten about her.
", "rawMarkdown": "Hi Andy. We'd love to have both you and Rashmi B on the team. Steven was caught in the severe winter storm so we're working on coordinating the invites now. Thank you, again!\n\nUpdate: seems only 1 invite can go at a time; I will contact Rashmi B to let her know we haven't forgotten about her.", "votes": null }, { "id": "265217", "postDate": "01/04/2018 21:52:42", "content": "Hi Andy. I've started with popular LGBM model few days ago, extend it with some features and other improvments, and now I near you. I think combination of different ideas may be useful.
", "rawMarkdown": "Hi Andy. I've started with popular LGBM model few days ago, extend it with some features and other improvments, and now I near you. I think combination of different ideas may be useful.", "votes": null }, { "id": "265220", "postDate": "01/04/2018 22:02:45", "content": "Thank you so much, Looking forward to working with you all.
", "rawMarkdown": "Thank you so much, Looking forward to working with you all.", "votes": null }, { "id": "265451", "postDate": "01/05/2018 15:01:57", "content": "\n\n\nunofficially in public on Instacart, where teaming wasn't allowed)
\n
Wow! Interesting that you publicly acknowledge some rule violation ...
", "rawMarkdown": "> unofficially in public on Instacart, where teaming wasn't allowed)\n\nWow! Interesting that you publicly acknowledge some rule violation ...", "votes": null }, { "id": "265462", "postDate": "01/05/2018 15:26:16", "content": "It's not rule violation. We interacted in public, in the discussion forums and public kernels, which was consistent with the rules. The rule is no private sharing of code or results, and there was none. But there are various degrees of cooperation within the public context.
", "rawMarkdown": "It's not rule violation. We interacted in public, in the discussion forums and public kernels, which was consistent with the rules. The rule is no private sharing of code or results, and there was none. But there are various degrees of cooperation within the public context.", "votes": null }, { "id": "265562", "postDate": "01/05/2018 21:23:22", "content": "Fair enough. ;)
", "rawMarkdown": "Fair enough. ;)", "votes": null }, { "id": "265895", "postDate": "01/07/2018 02:52:11", "content": "Whatever happened to that contest? I don't remember them publicly hiring anyone from it.
", "rawMarkdown": "Whatever happened to that contest? I don't remember them publicly hiring anyone from it.", "votes": null }, { "id": "265934", "postDate": "01/07/2018 05:51:26", "content": "Hi Andy, I would be willing to join teams. please send me an email at om.suchak@gmail.com if you are interested. cheers, Om
", "rawMarkdown": "Hi Andy, I would be willing to join teams. please send me an email at om.suchak@gmail.com if you are interested. cheers, Om", "votes": null } ], "comments": [ { "id": 264769, "author_name": "thrustmaster99", "author_url": "", "post_date": "01/03/2018 20:35:22", "content": "Hi Andy, I have achieved a score of 0.512 with different models tried out. I have few more models to try and check. It will be great if we can form a team and share our ideas with each other.
\n\nLet me know.
\n\nThanks.
", "votes": null, "replies": [] }, { "id": 264891, "author_name": "npa02012", "author_url": "", "post_date": "01/04/2018 03:38:57", "content": "Hey Andy. I may be interested, or if there is anyone else out there interested in talking. I don't want to get into detail publicly, but I have a .510 using the LGBM framework. I work in R and use data.table. Models use essentially the same framework as LGBM Starter.
\n\nI don't have a team friendly coding framework, however, I think sharing FE ideas, and some 'novel' framework ideas could be very beneficial.
\n\nI will be able to reply fast and share ideas quickly through the final submission.
\n\nEdit: Found a team.
", "votes": null, "replies": [] }, { "id": 265130, "author_name": "stevenknguyen", "author_url": "", "post_date": "01/04/2018 17:27:31", "content": "People think I'm always trolling them, and they're probably right. If you can deal with that would you like to join our team?
", "votes": null, "replies": [] }, { "id": 265147, "author_name": "learnmower", "author_url": "", "post_date": "01/04/2018 18:09:06", "content": "Hi Andy,\nIt would be an honor to work with you on this competition. If you haven't found a team, yet, then please consider Steven's invite. He's actually a very cordial person - I haven't been trolled, yet - and he's made my on-boarding to the team pleasant and smooth. We are using a couple of services to help with communicating ideas. The teammates are all very respectful, positive and hardworking. We hope you will consider our invite.
\n\nThank you.
", "votes": null, "replies": [ { "id": 265158, "author_name": "aharless", "author_url": "", "post_date": "01/04/2018 18:52:46", "content": "Yes, I've worked with Steven before (on Wikipedia -- and unofficially in public on Instacart, where teaming wasn't allowed). This does seem like a promising possibility. I already have an email reply out to RB (currently #63), who was the first to respond to my post. (I tried to send her a merge request, but it didn't work: seems there is a bug with very short team names.)
\n\nI am particularly interested if anyone has a method for predicting sales of new items. What I have tried so far hasn't worked at all, except for just using a small constant, but I haven't yet tried the obvious things that require a little more database work.
", "votes": null, "replies": [] }, { "id": 265164, "author_name": "aharless", "author_url": "", "post_date": "01/04/2018 19:04:10", "content": "OK, RB has changed her name to \"Rashmi B\", which works with merge requests. Also she sent me a merge request. What do you guys think about a team of 6?
", "votes": null, "replies": [] }, { "id": 265188, "author_name": "learnmower", "author_url": "", "post_date": "01/04/2018 20:21:29", "content": "RB will be a lucky person to have you as a teammate. I'm glad that the issues were corrected. I am open to more the merrier. Let me take this back to the team and get their thoughts. Thank you, again, for keeping open the opportunity to collaborate.
", "votes": null, "replies": [] }, { "id": 265202, "author_name": "learnmower", "author_url": "", "post_date": "01/04/2018 21:09:35", "content": "Hi Andy. We'd love to have both you and Rashmi B on the team. Steven was caught in the severe winter storm so we're working on coordinating the invites now. Thank you, again!
\n\nUpdate: seems only 1 invite can go at a time; I will contact Rashmi B to let her know we haven't forgotten about her.
", "votes": null, "replies": [] }, { "id": 265220, "author_name": "rashmibanthia", "author_url": "", "post_date": "01/04/2018 22:02:45", "content": "Thank you so much, Looking forward to working with you all.
", "votes": null, "replies": [] }, { "id": 265451, "author_name": "cpmpml", "author_url": "", "post_date": "01/05/2018 15:01:57", "content": "\n\n\nunofficially in public on Instacart, where teaming wasn't allowed)
\n
Wow! Interesting that you publicly acknowledge some rule violation ...
", "votes": null, "replies": [] }, { "id": 265462, "author_name": "aharless", "author_url": "", "post_date": "01/05/2018 15:26:16", "content": "It's not rule violation. We interacted in public, in the discussion forums and public kernels, which was consistent with the rules. The rule is no private sharing of code or results, and there was none. But there are various degrees of cooperation within the public context.
", "votes": null, "replies": [] }, { "id": 265562, "author_name": "cpmpml", "author_url": "", "post_date": "01/05/2018 21:23:22", "content": "Fair enough. ;)
", "votes": null, "replies": [] }, { "id": 265895, "author_name": "stevenknguyen", "author_url": "", "post_date": "01/07/2018 02:52:11", "content": "Whatever happened to that contest? I don't remember them publicly hiring anyone from it.
", "votes": null, "replies": [] } ] }, { "id": 265217, "author_name": "rekcahd", "author_url": "", "post_date": "01/04/2018 21:52:42", "content": "Hi Andy. I've started with popular LGBM model few days ago, extend it with some features and other improvments, and now I near you. I think combination of different ideas may be useful.
", "votes": null, "replies": [] }, { "id": 265934, "author_name": "omsuchak", "author_url": "", "post_date": "01/07/2018 05:51:26", "content": "Hi Andy, I would be willing to join teams. please send me an email at om.suchak@gmail.com if you are interested. cheers, Om
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "264739": "There isn't a lot of time left to try out new ideas, and there's still a lot of tuning to do on the old ideas. But maybe your old ideas are a little different from mine, and they will be useful to combine? In any case, as far as tuning goes, two computers are better than one, and N*K computers are better than K (where K is the average number of computers per teammate and N is the number of teammates).\n\nWhat I have now is mostly based on close variations of the frameworks and feature sets in public kernels, with some tuning and swapping of methods applied (and with some attempts to improve validation and to test robustness with respect to the public/private split, and with some mostly unsuccessful testing of additional features). So I'd be particularly interested in teammates that have had some success with other feature engineering or with frameworks other than the one popularized by Ceshine Lee.\n\nAnyhow, if you have reason to believe we can be a whole that is greater than the sum of its parts, let me know.", "264769": "Hi Andy, I have achieved a score of 0.512 with different models tried out. I have few more models to try and check. It will be great if we can form a team and share our ideas with each other. \n\nLet me know. \n\nThanks.", "264891": "Hey Andy. I may be interested, or if there is anyone else out there interested in talking. I don't want to get into detail publicly, but I have a .510 using the LGBM framework. I work in R and use data.table. Models use essentially the same framework as LGBM Starter. \n \nI don't have a team friendly coding framework, however, I think sharing FE ideas, and some 'novel' framework ideas could be very beneficial. \n \nI will be able to reply fast and share ideas quickly through the final submission.\n\nEdit: Found a team.", "265130": "People think I'm always trolling them, and they're probably right. If you can deal with that would you like to join our team?", "265147": "Hi Andy,\nIt would be an honor to work with you on this competition. If you haven't found a team, yet, then please consider Steven's invite. He's actually a very cordial person - I haven't been trolled, yet - and he's made my on-boarding to the team pleasant and smooth. We are using a couple of services to help with communicating ideas. The teammates are all very respectful, positive and hardworking. We hope you will consider our invite. \n\nThank you.", "265158": "Yes, I've worked with Steven before (on Wikipedia -- and unofficially in public on Instacart, where teaming wasn't allowed). This does seem like a promising possibility. I already have an email reply out to RB (currently #63), who was the first to respond to my post. (I tried to send her a merge request, but it didn't work: seems there is a bug with very short team names.)\n\nI am particularly interested if anyone has a method for predicting sales of new items. What I have tried so far hasn't worked at all, except for just using a small constant, but I haven't yet tried the obvious things that require a little more database work.", "265164": "OK, RB has changed her name to \"Rashmi B\", which works with merge requests. Also she sent me a merge request. What do you guys think about a team of 6?", "265188": "RB will be a lucky person to have you as a teammate. I'm glad that the issues were corrected. I am open to more the merrier. Let me take this back to the team and get their thoughts. Thank you, again, for keeping open the opportunity to collaborate.", "265202": "Hi Andy. We'd love to have both you and Rashmi B on the team. Steven was caught in the severe winter storm so we're working on coordinating the invites now. Thank you, again!\n\nUpdate: seems only 1 invite can go at a time; I will contact Rashmi B to let her know we haven't forgotten about her.", "265217": "Hi Andy. I've started with popular LGBM model few days ago, extend it with some features and other improvments, and now I near you. I think combination of different ideas may be useful.", "265220": "Thank you so much, Looking forward to working with you all.", "265451": "> unofficially in public on Instacart, where teaming wasn't allowed)\n\nWow! Interesting that you publicly acknowledge some rule violation ...", "265462": "It's not rule violation. We interacted in public, in the discussion forums and public kernels, which was consistent with the rules. The rule is no private sharing of code or results, and there was none. But there are various degrees of cooperation within the public context.", "265562": "Fair enough. ;)", "265895": "Whatever happened to that contest? I don't remember them publicly hiring anyone from it.", "265934": "Hi Andy, I would be willing to join teams. please send me an email at om.suchak@gmail.com if you are interested. cheers, Om" }, "source": "meta" }