Currently I have a NN with lb 2186, and a LGM with lb 2205. The details of my NN model can be found in the Best Single Model thread. I'm still working on the NN model, and it may have some room to boost.
\n\nI want to join a team with LGM model below 2200, hoping blend can make more boost. I can communiate with English and Chinese.
\n\nPlease send me a PM.
", "messages": [ { "id": 344115, "postDate": "2018-06-17T02:45:49.957Z", "content": "Currently I have a NN with lb 2186, and a LGM with lb 2205. The details of my NN model can be found in the Best Single Model thread. I'm still working on the NN model, and it may have some room to boost.
\n\nI want to join a team with LGM model below 2200, hoping blend can make more boost. I can communiate with English and Chinese.
\n\nPlease send me a PM.
", "rawMarkdown": "Currently I have a NN with lb 2186, and a LGM with lb 2205. The details of my NN model can be found in the [Best Single Model](https://www.kaggle.com/c/avito-demand-prediction/discussion/56816) thread. I'm still working on the NN model, and it may have some room to boost. \n\nI want to join a team with LGM model below 2200, hoping blend can make more boost. I can communiate with English and Chinese.\n\nPlease send me a PM.", "votes": 3 }, { "id": 345236, "postDate": "2018-06-19T13:12:20.207Z", "content": "I am not sure if you are looking for more features, but here is what I have.\nImage features: Distributions, Quality Metrics, Contours Distributions\nText POS : Parts of speech features\nTrend Features: When an ad is posted, what are other ads are being displayed across category, subcategory, region and city.\nText Probability Class: Divided the probability to four bins, and got the probabilities for each item_id. This becomes input features
\n\nIf you think any of these features might help, I will send you request. Total feature count is around 500.
\n\nI havent uploaded my results, actually I still did not build a model, as I dont find time to.
", "rawMarkdown": "I am not sure if you are looking for more features, but here is what I have.\nImage features: Distributions, Quality Metrics, Contours Distributions\nText POS : Parts of speech features\nTrend Features: When an ad is posted, what are other ads are being displayed across category, subcategory, region and city.\nText Probability Class: Divided the probability to four bins, and got the probabilities for each item_id. This becomes input features\n\nIf you think any of these features might help, I will send you request. Total feature count is around 500.\n\nI havent uploaded my results, actually I still did not build a model, as I dont find time to." }, { "id": 344539, "postDate": "2018-06-18T07:09:53.320Z", "content": "@LIu Jilong: I sent you a PM. As an update, our team now has multiple single model LGBs at 0.219X LB and we're working them into our blend as we speak. Your NN could definitely help us all get gold medals.
", "rawMarkdown": "@LIu Jilong: I sent you a PM. As an update, our team now has multiple single model LGBs at 0.219X LB and we're working them into our blend as we speak. Your NN could definitely help us all get gold medals.", "replies": [ { "id": 344661, "postDate": "2018-06-18T13:27:24.060Z", "content": "Maybe we should merge our teams into
\n\nghost, dizzl, oldguy, cat, Matt & 4 geese
\n\n:D
", "rawMarkdown": "Maybe we should merge our teams into\n\nghost, dizzl, oldguy, cat, Matt & 4 geese\n\n:D" }, { "id": 344688, "postDate": "2018-06-18T14:15:16.983Z", "content": "D:
", "rawMarkdown": "D:" }, { "id": 344761, "postDate": "2018-06-18T16:34:27.043Z", "content": "@Dieter: Not sure if you're serious or not but I love that team name. :)
", "rawMarkdown": "@Dieter: Not sure if you're serious or not but I love that team name. :)" }, { "id": 345151, "postDate": "2018-06-19T09:25:19.610Z", "content": "Sorry for late reply, I do not find the PM in kaggle's Team tab. Is there any problem?
", "rawMarkdown": "Sorry for late reply, I do not find the PM in kaggle's Team tab. Is there any problem?" }, { "id": 345153, "postDate": "2018-06-19T09:29:39.707Z", "content": "You can check your email address you used to register kaggle account
", "rawMarkdown": "You can check your email address you used to register kaggle account " }, { "id": 345333, "postDate": "2018-06-19T17:33:09.353Z", "content": "Hey @Liu Jilong I sent you another email. Just wanted to mention it here also in case you don't see it again.
", "rawMarkdown": "Hey @Liu Jilong I sent you another email. Just wanted to mention it here also in case you don't see it again." } ] }, { "id": 344147, "postDate": "2018-06-17T04:55:56.070Z", "content": "Hi, Liu. We have a strong LGB and a weak NN, I have sent a PM to you.
", "rawMarkdown": "Hi, Liu. We have a strong LGB and a weak NN, I have sent a PM to you." }, { "id": 344131, "postDate": "2018-06-17T03:40:05.363Z", "content": "Hi, Liu. Do you want join our team? we have strong single model and stacking here!!
", "rawMarkdown": "Hi, Liu. Do you want join our team? we have strong single model and stacking here!!", "replies": [ { "id": 344282, "postDate": "2018-06-17T14:06:17.110Z", "content": "May I know what your strong model is, which kinds of feature do you use?
", "rawMarkdown": "May I know what your strong model is, which kinds of feature do you use?" }, { "id": 344298, "postDate": "2018-06-17T14:51:36.680Z", "content": "Our strongest individual model(LGBM) only achieve cv 0.2148, but stacking take us below 0.213. We just apply label/likelihood encoding to different categorical, basic text feature, applicable text feature, stat image feature, n-way interaction, some binning to numeric and some stuff from Mercari. Currently we lack strong nn for our ensemble.
", "rawMarkdown": "Our strongest individual model(LGBM) only achieve cv 0.2148, but stacking take us below 0.213. We just apply label/likelihood encoding to different categorical, basic text feature, applicable text feature, stat image feature, n-way interaction, some binning to numeric and some stuff from Mercari. Currently we lack strong nn for our ensemble." }, { "id": 344501, "postDate": "2018-06-18T05:36:12.343Z", "content": "Hi! I am a teammate with Shuo Jen, Chang.\nWe have strong LGBM, XGB models(LB around 0.2190). But our NN is weak (LB is 0.2210~2220).\nAnd we have tricky stacking solution. So our score is so good. If you merge our team, your NN boost our score to gold line!\nPlease join our team!
", "rawMarkdown": "Hi! I am a teammate with Shuo Jen, Chang.\nWe have strong LGBM, XGB models(LB around 0.2190). But our NN is weak (LB is 0.2210~2220).\nAnd we have tricky stacking solution. So our score is so good. If you merge our team, your NN boost our score to gold line!\nPlease join our team!" }, { "id": 344534, "postDate": "2018-06-18T06:58:57.560Z", "content": "Please send me a PM.
", "rawMarkdown": "Please send me a PM." }, { "id": 344541, "postDate": "2018-06-18T07:14:16.900Z", "content": "Sure, I sent you a PM, have a look!
", "rawMarkdown": "Sure, I sent you a PM, have a look!" }, { "id": 345082, "postDate": "2018-06-19T07:05:12.297Z", "content": "@liu\nHi! Please check our PM and reply. Team merge deadline is coming soon.
", "rawMarkdown": "@liu\nHi! Please check our PM and reply. Team merge deadline is coming soon." }, { "id": 345150, "postDate": "2018-06-19T09:24:52.277Z", "content": "Sorry for late reply, this is the first time to team up with others. I do not find the PM in kaggle's Team tab. Is there any problem?
", "rawMarkdown": "Sorry for late reply, this is the first time to team up with others. I do not find the PM in kaggle's Team tab. Is there any problem?" }, { "id": 345154, "postDate": "2018-06-19T09:33:44.520Z", "content": "oh, check the email when you registered kaggle or you directly use \"contact user\" tab in my page to contact me.
", "rawMarkdown": "oh, check the email when you registered kaggle or you directly use \"contact user\" tab in my page to contact me." }, { "id": 345200, "postDate": "2018-06-19T11:51:26.297Z", "content": "please send me a team merge request directly, and i will make my decision tonight.
", "rawMarkdown": "please send me a team merge request directly, and i will make my decision tonight." }, { "id": 345220, "postDate": "2018-06-19T12:27:06.850Z", "content": "I can't send you team merge tonight since the sum of your entries and our entries have reached the maximum until today(285), we can only send the team merge request tomorrow(290). sorry for that.
", "rawMarkdown": "I can't send you team merge tonight since the sum of your entries and our entries have reached the maximum until today(285), we can only send the team merge request tomorrow(290). sorry for that. " }, { "id": 345281, "postDate": "2018-06-19T15:07:41.920Z", "content": "Please check your email (kaggle account).
", "rawMarkdown": "Please check your email (kaggle account)." }, { "id": 345325, "postDate": "2018-06-19T17:22:19.677Z", "rawMarkdown": "", "isDeleted": true } ] } ], "comments": [ { "id": 345236, "author_name": "ai sandwich", "author_url": "", "post_date": "2018-06-19T13:12:20.207000", "content": "I am not sure if you are looking for more features, but here is what I have.\nImage features: Distributions, Quality Metrics, Contours Distributions\nText POS : Parts of speech features\nTrend Features: When an ad is posted, what are other ads are being displayed across category, subcategory, region and city.\nText Probability Class: Divided the probability to four bins, and got the probabilities for each item_id. This becomes input features
\n\nIf you think any of these features might help, I will send you request. Total feature count is around 500.
\n\nI havent uploaded my results, actually I still did not build a model, as I dont find time to.
", "votes": 0, "replies": [] }, { "id": 344539, "author_name": "Peter Hurford", "author_url": "", "post_date": "2018-06-18T07:09:53.320000", "content": "@LIu Jilong: I sent you a PM. As an update, our team now has multiple single model LGBs at 0.219X LB and we're working them into our blend as we speak. Your NN could definitely help us all get gold medals.
", "votes": 0, "replies": [ { "id": 344661, "author_name": "Dieter", "author_url": "", "post_date": "2018-06-18T13:27:24.060000", "content": "Maybe we should merge our teams into
\n\nghost, dizzl, oldguy, cat, Matt & 4 geese
\n\n:D
", "votes": 0, "replies": [] }, { "id": 344688, "author_name": "Angus Chang", "author_url": "", "post_date": "2018-06-18T14:15:16.983000", "content": "D:
", "votes": 0, "replies": [] }, { "id": 344761, "author_name": "Peter Hurford", "author_url": "", "post_date": "2018-06-18T16:34:27.043000", "content": "@Dieter: Not sure if you're serious or not but I love that team name. :)
", "votes": 0, "replies": [] }, { "id": 345151, "author_name": "Liu Jilong", "author_url": "", "post_date": "2018-06-19T09:25:19.610000", "content": "Sorry for late reply, I do not find the PM in kaggle's Team tab. Is there any problem?
", "votes": 0, "replies": [] }, { "id": 345153, "author_name": "yimacs", "author_url": "", "post_date": "2018-06-19T09:29:39.707000", "content": "You can check your email address you used to register kaggle account
", "votes": 0, "replies": [] }, { "id": 345333, "author_name": "Peter Hurford", "author_url": "", "post_date": "2018-06-19T17:33:09.353000", "content": "Hey @Liu Jilong I sent you another email. Just wanted to mention it here also in case you don't see it again.
", "votes": 0, "replies": [] } ] }, { "id": 344147, "author_name": "zr", "author_url": "", "post_date": "2018-06-17T04:55:56.070000", "content": "Hi, Liu. We have a strong LGB and a weak NN, I have sent a PM to you.
", "votes": 0, "replies": [] }, { "id": 344131, "author_name": "Angus Chang", "author_url": "", "post_date": "2018-06-17T03:40:05.363000", "content": "Hi, Liu. Do you want join our team? we have strong single model and stacking here!!
", "votes": 0, "replies": [ { "id": 344282, "author_name": "Liu Jilong", "author_url": "", "post_date": "2018-06-17T14:06:17.110000", "content": "May I know what your strong model is, which kinds of feature do you use?
", "votes": 0, "replies": [] }, { "id": 344298, "author_name": "Angus Chang", "author_url": "", "post_date": "2018-06-17T14:51:36.680000", "content": "Our strongest individual model(LGBM) only achieve cv 0.2148, but stacking take us below 0.213. We just apply label/likelihood encoding to different categorical, basic text feature, applicable text feature, stat image feature, n-way interaction, some binning to numeric and some stuff from Mercari. Currently we lack strong nn for our ensemble.
", "votes": 0, "replies": [] }, { "id": 344501, "author_name": "takuoko", "author_url": "", "post_date": "2018-06-18T05:36:12.343000", "content": "Hi! I am a teammate with Shuo Jen, Chang.\nWe have strong LGBM, XGB models(LB around 0.2190). But our NN is weak (LB is 0.2210~2220).\nAnd we have tricky stacking solution. So our score is so good. If you merge our team, your NN boost our score to gold line!\nPlease join our team!
", "votes": 0, "replies": [] }, { "id": 344534, "author_name": "Liu Jilong", "author_url": "", "post_date": "2018-06-18T06:58:57.560000", "content": "Please send me a PM.
", "votes": 0, "replies": [] }, { "id": 344541, "author_name": "Angus Chang", "author_url": "", "post_date": "2018-06-18T07:14:16.900000", "content": "Sure, I sent you a PM, have a look!
", "votes": 0, "replies": [] }, { "id": 345082, "author_name": "takuoko", "author_url": "", "post_date": "2018-06-19T07:05:12.297000", "content": "@liu\nHi! Please check our PM and reply. Team merge deadline is coming soon.
", "votes": 0, "replies": [] }, { "id": 345150, "author_name": "Liu Jilong", "author_url": "", "post_date": "2018-06-19T09:24:52.277000", "content": "Sorry for late reply, this is the first time to team up with others. I do not find the PM in kaggle's Team tab. Is there any problem?
", "votes": 0, "replies": [] }, { "id": 345154, "author_name": "Angus Chang", "author_url": "", "post_date": "2018-06-19T09:33:44.520000", "content": "oh, check the email when you registered kaggle or you directly use \"contact user\" tab in my page to contact me.
", "votes": 0, "replies": [] }, { "id": 345200, "author_name": "Liu Jilong", "author_url": "", "post_date": "2018-06-19T11:51:26.297000", "content": "please send me a team merge request directly, and i will make my decision tonight.
", "votes": 0, "replies": [] }, { "id": 345220, "author_name": "Angus Chang", "author_url": "", "post_date": "2018-06-19T12:27:06.850000", "content": "I can't send you team merge tonight since the sum of your entries and our entries have reached the maximum until today(285), we can only send the team merge request tomorrow(290). sorry for that.
", "votes": 0, "replies": [] }, { "id": 345281, "author_name": "YoungLamb", "author_url": "", "post_date": "2018-06-19T15:07:41.920000", "content": "Please check your email (kaggle account).
", "votes": 0, "replies": [] }, { "id": 345325, "author_name": "", "author_url": "", "post_date": "2018-06-19T17:22:19.677000", "content": "", "votes": 0, "replies": [] } ] } ], "raw_markdown_by_id": { "344115": "Currently I have a NN with lb 2186, and a LGM with lb 2205. The details of my NN model can be found in the [Best Single Model](https://www.kaggle.com/c/avito-demand-prediction/discussion/56816) thread. I'm still working on the NN model, and it may have some room to boost. \n\nI want to join a team with LGM model below 2200, hoping blend can make more boost. I can communiate with English and Chinese.\n\nPlease send me a PM.", "345236": "I am not sure if you are looking for more features, but here is what I have.\nImage features: Distributions, Quality Metrics, Contours Distributions\nText POS : Parts of speech features\nTrend Features: When an ad is posted, what are other ads are being displayed across category, subcategory, region and city.\nText Probability Class: Divided the probability to four bins, and got the probabilities for each item_id. This becomes input features\n\nIf you think any of these features might help, I will send you request. Total feature count is around 500.\n\nI havent uploaded my results, actually I still did not build a model, as I dont find time to.", "344539": "@LIu Jilong: I sent you a PM. As an update, our team now has multiple single model LGBs at 0.219X LB and we're working them into our blend as we speak. Your NN could definitely help us all get gold medals.", "344147": "Hi, Liu. We have a strong LGB and a weak NN, I have sent a PM to you.", "344131": "Hi, Liu. Do you want join our team? we have strong single model and stacking here!!" } }